USF Libraries
USF Digital Collections

Statistical and data mining methodologies for behavioral analysis in transgenic mouse models of Alzheimer's disease

MISSING IMAGE

Material Information

Title:
Statistical and data mining methodologies for behavioral analysis in transgenic mouse models of Alzheimer's disease parallels with human AD evaluation
Physical Description:
Book
Language:
English
Creator:
Leighty, Ralph E
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Neuropathology
Neural networks
Caffeine
GRK5
GMCSF
Dissertations, Academic -- Cellular Biology, Microbiology, and Molecular Biology -- Doctoral -- USF   ( lcsh )
Genre:
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: Alzheimer's disease (AD) is a progressive neurodegenerative disorder and the leading cause of human senile dementia. Alzheimer's represents a significant public health concern, having widespread social and economic implications. Consequently, protocols for early detection and therapeutic intervention (both behavioral and pharmacologic) constitute important targets for medical investigation. Furthermore, contemporary research depends upon comprehensive neurobehavioral assessment and advanced statistical and computational analytic methodologies for characterizing AD-associated sensorimotor and cognitive impairment, as well as evaluating therapeutic efficacy. This dissertation introduces data mining-based techniques (decision trees, neural networks, support vector machines) for behavioral analysis in both nontransgenic and Alzheimer's transgenic mice, to evaluate the cognitive benefits of long-term caffeine treatment.Both treatment and transgenic effects are identified through advanced statistical (discriminant analysis) and data mining approaches. In addition, a novel mouse-based cognitive assessment paradigm, adapted from a human interference learning AD-diagnostic protocol, is implemented to evaluate both genetic (GRK5) and therapeutic (GM-CSF) effects in mice, against an Alzheimer's transgenic background. Data mining techniques are shown to be comparable to conventional statistical analyses, often providing complementary diagnostic information. Indeed, comparisons between data mining-based and multivariate statistical analyses, with respect to groupwise discriminability, support the use of both methodologies in neurobehavioral research.Future work involving both data mining-based and multivariate statistical analyses of cognitive-behavioral data is discussed, emphasizing the need for longitudinal studies, repeated-measure designs, and spatiotemporal modeling for evaluating the time-course of both human AD and AD-like pathology in transgenic mouse models.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2009.
Bibliography:
Includes bibliographical references.
System Details:
Mode of access: World Wide Web.
System Details:
System requirements: World Wide Web browser and PDF reader.
Statement of Responsibility:
by Ralph E. Leighty.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 313 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002029068
oclc - 436864377
usfldc doi - E14-SFE0002869
usfldc handle - e14.2869
System ID:
SFS0027186:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 2200409Ka 4500
controlfield tag 001 002029068
005 20090916093728.0
007 cr mnu|||uuuuu
008 090916s2009 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002869
035
(OCoLC)436864377
040
FHM
c FHM
049
FHMM
090
QR41.2 (Online)
1 100
Leighty, Ralph E.
0 245
Statistical and data mining methodologies for behavioral analysis in transgenic mouse models of Alzheimer's disease :
b parallels with human AD evaluation
h [electronic resource] /
by Ralph E. Leighty.
260
[Tampa, Fla] :
University of South Florida,
2009.
500
Title from PDF of title page.
Document formatted into pages; contains 313 pages.
Includes vita.
502
Dissertation (Ph.D.)--University of South Florida, 2009.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
520
ABSTRACT: Alzheimer's disease (AD) is a progressive neurodegenerative disorder and the leading cause of human senile dementia. Alzheimer's represents a significant public health concern, having widespread social and economic implications. Consequently, protocols for early detection and therapeutic intervention (both behavioral and pharmacologic) constitute important targets for medical investigation. Furthermore, contemporary research depends upon comprehensive neurobehavioral assessment and advanced statistical and computational analytic methodologies for characterizing AD-associated sensorimotor and cognitive impairment, as well as evaluating therapeutic efficacy. This dissertation introduces data mining-based techniques (decision trees, neural networks, support vector machines) for behavioral analysis in both nontransgenic and Alzheimer's transgenic mice, to evaluate the cognitive benefits of long-term caffeine treatment.Both treatment and transgenic effects are identified through advanced statistical (discriminant analysis) and data mining approaches. In addition, a novel mouse-based cognitive assessment paradigm, adapted from a human interference learning AD-diagnostic protocol, is implemented to evaluate both genetic (GRK5) and therapeutic (GM-CSF) effects in mice, against an Alzheimer's transgenic background. Data mining techniques are shown to be comparable to conventional statistical analyses, often providing complementary diagnostic information. Indeed, comparisons between data mining-based and multivariate statistical analyses, with respect to groupwise discriminability, support the use of both methodologies in neurobehavioral research.Future work involving both data mining-based and multivariate statistical analyses of cognitive-behavioral data is discussed, emphasizing the need for longitudinal studies, repeated-measure designs, and spatiotemporal modeling for evaluating the time-course of both human AD and AD-like pathology in transgenic mouse models.
538
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
590
Co-advisor: Gary Arendash, Ph.D.
Co-advisor: Huntington Potter, Ph.D.
653
Neuropathology
Neural networks
Caffeine
GRK5
GMCSF
690
Dissertations, Academic
z USF
x Cellular Biology, Microbiology, and Molecular Biology
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.2869



PAGE 1

StatisticalandDataMiningMethodologiesforBehavioralAnalysisinTransgenicMouseModelsofAlzheimer'sDisease:ParallelswithHumanADEvaluationbyRalphE.LeightyAdissertationsubmittedinpartialfulllmentoftherequirementsforthedegreeofDoctorofPhilosophyDepartmentofCellularBiology,Microbiology,andMolecularBiologyCollegeofArtsandSciencesUniversityofSouthFloridaCo-MajorProfessor:GaryArendash,Ph.D.Co-MajorProfessor:HuntingtonPotter,Ph.D.GordonFox,Ph.D.PatrickBradshaw,Ph.D.ChuanhaiCao,Ph.D.DateofApproval:April6,2009Keywords:neuropathology,neuralnetworks,caeine,GRK5,GMCSFcCopyright2009,RalphE.Leighty

PAGE 2

TABLEOFCONTENTSLISTOFTABLESvLISTOFFIGURESviiABSTRACTviiiCHAPTER1BACKGROUND11.1HistoryofAlzheimer'sDisease...................11.2BehavioralCharacterizationofAlzheimer'sDisease.......21.3PathologicalCharacterizationofAlzheimer'sDisease......31.4DiagnosisofAlzheimer'sDisease..................61.4.1PsychometricAssessment................71.4.2Biomarkers........................111.4.3NeurologicalBasisofMemorySystems.........141.4.4Electroencephalography.................171.4.5DiagnosticMedicalImaging...............201.5RiskFactorsforAlzheimer'sDisease...............231.6RiskReductionStrategies.....................251.7GeneticsofAlzheimer'sDisease..................291.8AnimalModelsofAlzheimer'sDisease..............331.8.1PDAPPTransgenicMouseModel............351.8.2APPswTransgenicMouseModels............381.8.3APPsw+PS1TransgenicMouseModel.........421.8.4APPsw+PS1+TauTransgenicMouseModel......441.8.5AnimalModels:ACoda.................451.9TreatmentsforAlzheimer'sDisease................471.9.1Pharmaceuticals......................471.9.2NaturalProducts.....................501.9.3Behavior-BasedTherapies................521.10StatementofPurpose........................54CHAPTER2BEHAVIORALASSESSMENTINALZHEIMER'STRANS-GENICMICE562.1SensorimotorTasksandAssociatedMeasures...........572.1.1OpenFieldActivity...................57i

PAGE 3

2.1.2BalanceBeam.......................582.1.3StringAgility.......................602.2CognitiveTasksandAssociatedMeasures............602.2.1Y-Maze..........................602.2.2ElevatedPlusMaze....................612.2.3MorrisWaterMazeSubmergedPlatform.......632.2.4CircularPlatform.....................642.2.5PlatformRecognition...................652.2.6RadialArmWaterMaze.................662.3NewParadigmsandParallels:TalesofMiceandMen......67CHAPTER3BEHAVIORALDATAANALYSIS743.1OnBehavioralClassication....................743.2StatisticalAnalysis.........................803.2.1CorrelationAnalysis...................813.2.2FactorAnalysis......................843.2.3DiscriminantAnalysis..................893.3DataMiningAnalysis........................913.3.1DecisionTrees.......................933.3.2NeuralNetworks.....................953.3.3SupportVectorMachines.................1013.4PracticalComparisons.......................1033.5What'sPastIsPrologue".....................106CHAPTER4CAFFEINEADMINISTRATIONINNONTRANSGENICMICE1074.1Introduction.............................1074.2MaterialsandMethods.......................1104.3ResultsofStatisticalAnalyses...................1134.3.1CorrelationAnalysis...................1134.3.2FactorAnalysis......................1144.3.3DiscriminantAnalysis..................1174.4ResultsofDataMiningAnalyses.................1184.4.1DecisionTreeAnalysis..................1184.4.2NeuralNetworkAnalysis.................1184.4.3SupportVectorMachineAnalysis............1194.5Discussion..............................119CHAPTER5CAFFEINEADMINISTRATIONINALZHEIMER'STRANSGENICMICE1225.1Introduction.............................1225.2MaterialsandMethods.......................1265.3ResultsofStatisticalAnalyses...................127ii

PAGE 4

5.3.1CorrelationAnalysis...................1275.3.2FactorAnalysis......................1285.3.3DiscriminantAnalysis..................1295.4ResultsofDataMiningAnalyses.................1335.4.1DecisionTreeAnalysis..................1335.4.2NeuralNetworkAnalysis.................1345.4.3SupportVectorMachineAnalysis............1355.5Discussion..............................136CHAPTER6INTERFERENCETESTINGINHUMANS:ACOMPARI-SONOFSTATISTICALANDDATAMININGMETHODS1406.1Introduction.............................1406.2MaterialsandMethods.......................1426.3ResultsofStatisticalAnalyses...................1426.3.1StandardAnalysis....................1426.3.2CorrelationAnalysis...................1436.3.3DiscriminantAnalysis..................1456.4ResultsofDataMiningAnalyses.................1486.4.1DecisionTreeAnalysis..................1486.4.2NeuralNetworkAnalysis.................1496.4.3SupportVectorMachineAnalysis............1506.5Discussion..............................151CHAPTER7THEINTERFERENCETASK:ANOVELASSESSMENTPARADIGM1567.1Introduction.............................1567.2MaterialsandMethods.......................1597.3ResultsofStatisticalAnalyses:ComprehensiveTaskBattery..1627.3.1StandardBehavioralAnalysis..............1627.3.2CorrelationAnalysis...................1657.3.3FactorAnalysis......................1667.3.4DiscriminantAnalysis..................1697.4ResultsofDataMiningAnalyses:ComprehensiveTaskBattery1727.4.1DecisionTreeAnalysis..................1727.4.2NeuralNetworkAnalysis.................1747.4.3SupportVectorMachineAnalysis............1767.5ResultsofStatisticalAnalyses:InterferenceParadigm......1777.5.1StandardBehavioralAnalysis..............1777.5.2CorrelationAnalysis...................1797.5.3FactorAnalysis......................1817.5.4DiscriminantAnalysis..................1837.6ResultsofDataMiningAnalyses:InterferenceParadigm....1867.6.1DecisionTreeAnalysis..................186iii

PAGE 5

7.6.2NeuralNetworkAnalysis.................1897.6.3SupportVectorMachineAnalysis............1917.7Discussion..............................193CHAPTER8INTERFERENCETASK-BASEDTHERAPEUTICEVALU-ATIONOFGM-CSF2048.1Introduction.............................2048.2MaterialsandMethods.......................2058.3ResultsofStatisticalAnalyses...................2088.3.1StandardStatisticalAnalysis..............2088.3.2CorrelationAnalysis...................2128.3.3FactorAnalysis......................2148.3.4DiscriminantAnalysis..................2178.4ResultsofDataMiningAnalyses.................2228.4.1DecisionTreeAnalysis..................2228.4.2NeuralNetworkAnalysis.................2258.4.3SupportVectorMachineAnalysis............2298.5Discussion..............................232CHAPTER9CONCLUSIONS239REFERENCES247ABOUTTHEAUTHOREndPageiv

PAGE 6

LISTOFTABLESTable2.1Task-associatedbehavioralmeasuresusedinanalyses.....59Table4.1Correlationsbetweenbehavioralmeasuresinthenontransgeniccaeinestudy...........................115Table4.2Unrotatedfactorcomponentloadingsinthenontransgeniccaf-feinestudy.............................116Table4.3Classierperformancecomparisoninthenontransgeniccaeinestudy................................120Table5.1CorrelationsbetweenbehavioralmeasuresintheAlzheimer'stransgeniccaeinestudy.....................127Table5.2UnrotatedfactorcomponentloadingsintheAlzheimer'strans-geniccaeinestudy........................128Table5.3ClassierperformancecomparisonintheAlzheimer'stransgeniccaeinestudy...........................130Table6.1CorrelationsamongMMSEscoresandbehavioralmeasuresinthehumansemanticinterferenceprotocol............143Table6.2Classierperformancecomparisoninthehumansemanticinter-ferenceprotocol..........................146Table7.1GroupwisecontrastsforallbehavioraltaskmeasuresintheGRK5study............................164Table7.2CorrelationsbetweenbehavioralmeasuresofthecomprehensivetaskbatteryintheGRK5study.................167Table7.3Varimax-rotatedfactoranalysisofcomprehensivetaskbatterymeasuresintheGRK5study...................168Table7.4Classierperformancecomparisonusingcomprehensivetaskbat-terymeasuresintheGRK5study................169v

PAGE 7

Table7.5CorrelationsbetweenbehavioralmeasuresoftheinterferenceparadigmintheGRK5study...................179Table7.6Varimax-rotatedfactoranalysisofinterferenceparadigmmea-suresintheGRK5study.....................181Table7.7ClassierperformancecomparisonusinginterferenceparadigmmeasuresintheGRK5study...................182Table8.1CorrelationsbetweenbehavioralmeasuresoftheinterferenceparadigmintheGMCSFstudy.................212Table8.2Varimax-rotatedfactoranalysisofinterferenceparadigmmea-suresintheGMCSFstudy....................214Table8.3ClassierperformancecomparisonusinginterferenceparadigmmeasuresintheGMCSFstudy..................216vi

PAGE 8

LISTOFFIGURESFigure4.1Comparisonsbetweenuntreatedandcaeine-treatednontrans-genicmice..............................109Figure4.2Sampledecisiontreeclassiergeneratedinthenontransgeniccaf-feinestudy.............................118Figure5.1BehavioralassessmentinNTandTgmicereceivingcaeine..125Figure5.2Canonicalscoresplotofdiscriminantanalysisforallthreegroups132Figure6.1Groupwisecontrastsforallhumaninterferenceprotocolmeasures144Figure7.1GroupwisecontrastsforallinterferenceparadigmmeasuresintheGRK5study............................178Figure8.1Groupwisecomparisonoferror-scoresinRAWMpost-testtrialsT4andT5,byblocks,intheGMCSFstudy..........209Figure8.2GroupwisecontrastsforallinterferenceparadigmmeasuresintheGMCSFstudy...........................211vii

PAGE 9

STATISTICALANDDATAMININGMETHODOLOGIESFORBEHAVIORALANALYSISINTRANSGENICMOUSEMODELSOFALZHEIMER'SDISEASE:PARALLELSWITHHUMANADEVALUATIONRalphE.LeightyABSTRACTAlzheimer'sdiseaseADisaprogressiveneurodegenerativedisorderandtheleadingcauseofhumanseniledementia.Alzheimer'srepresentsasignicantpublichealthconcern,havingwidespreadsocialandeconomicimplications.Consequently,protocolsforearlydetectionandtherapeuticinterventionbothbehavioralandphar-macologicconstituteimportanttargetsformedicalinvestigation.Furthermore,con-temporaryresearchdependsuponcomprehensiveneurobehavioralassessmentandadvancedstatisticalandcomputationalanalyticmethodologiesforcharacterizingAD-associatedsensorimotorandcognitiveimpairment,aswellasevaluatingthera-peuticecacy.Thisdissertationintroducesdatamining-basedtechniquesdecisiontrees,neuralnetworks,supportvectormachinesforbehavioralanalysisinbothnontransgenicandAlzheimer'stransgenicmice,toevaluatethecognitivebenetsoflong-termcaeinetreatment.Bothtreatmentandtransgeniceectsareidentiedthroughadvancedstatisticaldiscriminantanalysisanddataminingapproaches.Inaddition,anovelmouse-basedcognitiveassessmentparadigm,adaptedfromahu-maninterferencelearningAD-diagnosticprotocol,isimplementedtoevaluatebothgeneticGRK5andtherapeuticGM-CSFeectsinmice,againstanAlzheimer'stransgenicbackground.Dataminingtechniquesareshowntobecomparabletocon-viii

PAGE 10

ventionalstatisticalanalyses,oftenprovidingcomplementarydiagnosticinformation.Indeed,comparisonsbetweendatamining-basedandmultivariatestatisticalanaly-ses,withrespecttogroupwisediscriminability,supporttheuseofbothmethodolo-giesinneurobehavioralresearch.Futureworkinvolvingbothdatamining-basedandmultivariatestatisticalanalysesofcognitive-behavioraldataisdiscussed,emphasiz-ingtheneedforlongitudinalstudies,repeated-measuredesigns,andspatiotemporalmodelingforevaluatingthetime-courseofbothhumanADandAD-likepathologyintransgenicmousemodels.ix

PAGE 11

CHAPTER1BACKGROUND1.1HistoryofAlzheimer'sDiseaseAlzheimer'sdiseaseAD;OMIM104300isaprogressive,degenerativeneurolog-icaldisorder,characterizedbysensorimotor,perceptual,andcognitiveimpairmentandthedevelopmentofdistinctiveneuropathologiclesions.Thedisorderwasrstdescribedin1907bytheGermanpsychiatristAloisAlzheimer1864-1915,whoreportedthecaseofa51-year-oldwomanwhoinitiallypresentedwithparanoia,confusion,andmemoryandlanguageimpairmentAlzheimer,1907;Stelzmannetal.,1995;Maureretal.,1997.Hermentaldeteriorationcontinueduntilherdeathfourandone-halfyearslater.Postmortemneurohistologicalexaminationrevealedproteinaceousextracellularneuriticplaquesandintracellularneurobrillarytangles;AlzheimerwasthersttoidentifythelatterndingGraeberetal.,1998;Graeber,1999.Thepresenceofbothneuropathologicmarkersrepresentsconrmatoryevi-denceofAlzheimer'sdisease,althoughneitherisspecicforADMarkesbery,1997;Selkoe,2001;AdlardandCummings,2004.TheleadingcauseofseniledementiaisAlzheimer'sdisease,aectingapproxi-matelytenpercentofpeopleovertheageof65yearsandfortypercentofthoseover80yearsofageEvansetal.,1989.Currently,aboutvemillionAmericanshaveAlzheimer'sdisease,andthenumbermayincreasetoasmanyassixteenmillionby2050Hebertetal.,2003;Alzheimer'sAssociation,2008.Theincreasingprevalence1

PAGE 12

ofAlzheimer'sdiseaserepresentsasignicantpublichealthconcernBrookmeyeretal.,1998,withwidespreadsocialandeconomicconsequences.1.2BehavioralCharacterizationofAlzheimer'sDiseaseReectingtheprogressiveneuropathology,thetimecourseofAlzheimer'sdiseasecorrespondstoagradeddeclineinmentalfunctionoccurringoveraperiodofbetweenveandtwentyyearsLocascioetal.,1995,whichresemblesaretrogressionofnor-maldevelopmentReisbergetal.,1999.Alzheimer'sdiseasehasbeencalleddeath...byathousandsubtractions"Shenk,2001inrecognitionofthisinsidiousand,ultimately,fataltrajectoryofattrition.Overtbehavioralincludingcognitiveman-ifestationsreecttheextentandprogressoftheunderlyingneuropathologyWeineretal.,2005.Alzheimer'sdisease,intheearlystage,ismarkedbysignicantimpairmentinshort-termworkingmemoryMcKhannetal.,1984;ForstlandKurtz,1999.Thisconditiontranscendscasualforgetfulnessorabsentmindedness"indailyroutine.Pa-tientsbeginexperiencingdicultywithtasksrequiringconcentrationandsustainedattention,planningandorganization.Psychiatricsymptoms,includingdepression,delusions,andanxiety,mayappeartowardtheendofthisstageHartetal.,2003andpersistthroughoutthepatient'slife.Thesedecitsimpedetheperformanceofdailyactivities,andcontributetoincreasedaccidentandhazardriskstopatients.InthemoderatestageofAlzheimer'sdisease,furtherdeclinesinshort-termmem-oryandthinkingskillscompromisethepatient'sabilitytocareforhim-orherself.Erosionoftheintegrityoflong-termmemoryleadstofurtherdisorientationandconfusion.Thisconfusion,inturn,causesagitation,anxiety,restlessness,andag-gression.Delusionsandhallucinationsofincreasingseverity,compulsivewanderingandhoardingbehaviorsareobservedaswellDevanandetal.,1997.2

PAGE 13

Nearlycompletelossofmentalfunction,inadditiontoprofoundmotorimpair-ment,characterizestheadvancedstageofAlzheimer'sdisease.Thesepatientsareinextremelyfragilehealth,oftenbecomingbedriddenandsinkingintoaminimally-responsivevegetativestatebeforenallysuccumbingtosecondaryillnessese.g.,respiratoryand/orcardiovascularpathology,opportunisticinfectionSourenetal.,1995.AfterclinicaldiagnosisofAD,theaveragedurationofpatientsurvivalisbetweenveandeightyearsBraccoetal.,1994.1.3PathologicalCharacterizationofAlzheimer'sDiseaseBeforeAlzheimerdescribedtheassociationbetweenprogressivedementiaandcorticallesions,manyage-associatedcerebrovascularstructuralanomalieshadbeenreportedinpatients.However,itwasunclearwhetherthelesionswerethecauseofthedementia.ElucidationoftheunderlyingmechanismsofAlzheimer'sdiseasedevelopedfromseverallinesofinquiryintotheneurophysiologicalcorrelatesoftheprogressivebehavioralmainlycognitivedysfunctionobservedinADpatients.OriginallydescribedbyAlzheimerin1907,thedistinguishingneuropathologicalmarkersofADareintercellularsenileplaquesandintracellularneurobrillarytan-glesNFTse.g.,Mattson,2004.Theplaquesarecomprisedofvariable-lengthto42residuesbetaamyloidproteinRoheretal.,1986andthetanglesconsistofhyperphosphorylatedcytoskeletaltauproteinGrundke-Iqbaletal.,1986.Thesefeaturesarefoundthroughoutthecerebralcortex,amygdala,andhippocampus.Withinthesespecicallyvulnerablebrainregions,synapticdisruptionandneuronalcelllossoccure.g.,Sternetal.,2004,withconcomitantneurochemicalchanges.TheunderlyingmechanismforthepathogenesisofAlzheimer'sdiseaseemphasizestheroleofbetaamyloidproteinHardy,1997;TanziandBertram,2005;Hardy,2006,identiedasthecorepeptideinplaquesGlennerandWong,1984a,1984b;3

PAGE 14

Mastersetal.,1985.ThisproteinappearshighlyconservedacrossspeciesSelkoeetal.,1987;antibodiestoapartialpeptidefragment8aminoacidsofhumanamyloidcross-reactedwithcerebrovasculardepositsandneuriticplaquesfromagedmammalsdog,monkey,orangutan,andpolarbear.ItisthroughtheaccumulationofabnormalbetaamyloidproteinthatAlzheimersdiseaseisconsideredaproteinmisfoldingdiseaseCohenandKelly,2003;Hashimotoetal.,2003,sharingaggrega-tionkineticspropertieswithpriondiseasese.g.,scrapie,Creutzfeldt-Jakobdisease,bovinespongiformencephalopathyGrith,1967;Comeetal.,1993;Cohen,1999;Hayashietal.,2004.Withprogressivelossofcorticalandhippocampalneurons,concomitantbrainatrophyisobservedinADbrains.Thecerebralcorticalatrophyishemisphericallyasymmetric,beingmorepronouncedinthelefthemispherethanintherightGeeetal.,2003;Thompsonetal.,2003,particularlyintheanterior-andposterolateral-temporalanddorsolateral-prefrontalregionsGeeetal.,2003.Thecerebellum,occipitalregion,andsensorimotorcortexarelargelysparedThompsonetal.,2003.InterruptionofconnectivitybetweenassociationcorticesmayarisefromthelossofpyramidalneuronsoftheCA1,subiculum,andentorhinalcortexofthehippocam-pusMorrisonandHof,1997.Withinthehippocampus,substantiallossofCA1neuronsisobservedinmildtosevereADpatientsPriceetal.,2001.Lossofhip-pocampalneuronsintheCA1andsubiculumareassociatedwiththeformationofneurobrillarytanglesRossleretal.,2002.Betaamyloidisproducedthroughaseriesofprotease-mediatedcleavagesfromamyloidprecursorproteinAPP;Goldgaberetal.,1987intoseveralisoformsofbetween365and750residuesVassaretal.,1999.APPisatransmembranepro-teinKangetal.,1987,havingalongextracellularamino-terminusandashorterintracellularcarboxy-terminusMaccionietal.,2001.APPisencodedbyagene4

PAGE 15

withintheDown'ssyndromeregionofchromosome21Tanzietal.,1987.Normally,alpha-secretasecleavesAPPnearthecenterofthebetaamyloiddomain,resultinginanon-amyloidogenicpolypeptidefragmentSelkoe,2001.However,twopathogenicformsofbetaamyloidA,betaamyloid[1-40]A40andbetaamyloid[1-42]A42,resultfromcleavageofAPPbybeta-secretaseatMet671followedbygamma-secretaseatVal711orIle713,respectivelyCitronetal.,1996;Bossy-Wetzeletal.,2004.Mis-sensemutationsencodedinAPPnearsecretasecleavagesitesleadtoincreasedbeta-amyloidproductionSelkoe,2001.NeitherAPPnorAisspecicforAlzheimersdisease;amyloidprecursorproteinandbetaamyloidarefoundinADpatientsaswellasnon-ADindividualsSelkoe,2001.RegiospecicaccumulationofA42resultsindiuseplaques,whichoccurmainlywithintheassociationandlimbiccorticesRoheretal.,2000.Monomericbetaamy-loidformalpha-heliceswhichundergodestabilizationandconformaltransformationintobeta-helicaldimersSerpell,2000,eachconsistingofahydrophobiccoresur-roundedbyhydrophilicresidues.Adjacentdimersbindtogetherintoprotolamentswhich,inturn,formbeta-sheets,andsubsequentlycoalesceintobrilsRoheretal.,2000;WalshandSelkoe,2004.Compactdenseplaquessubsequentlymayformfromadditionalconformalchangesinbrilaggregatese.g.,Kayedetal.,2003,andtriggerastrocyticandmicroglialactivationYamaguchi,etal.,1988,whichrepre-sentaninammatoryresponseSelkoe,2001.Subsequentproductionoffreeradicalsbymicroglialmitochondrialeadstoneuronaldysfunctionandcelldeaththroughox-idativedamage,suchaslipidperoxidationSelkoe,2001;Reddy,2006.Astrocytesproducetwoinammatoryproteins{-antichymotrypsinACTandapolipoproteinEApoE{whichparticipateinamyloidplaqueformationSelkoe,2001;Potteretal.,2001.BothACTandApoEhavebeenshowntopromoteamyloidforma-tionanddepositionaspathologicchaperones"bothinvitroandinvivoSanan5

PAGE 16

etal.,1994;Wisniewskietal.,1994;Balesetal.,1999.Furthermore,ithasbeendemonstratedthatACTandApoEacteithertogetherorindependentlytopromotebothdiuseandcompactbetaamyloidplaqueswithoutinuencingmonomericbetaamyloidlevelsNilssonetal.,2004.IntracellularNFTsarecomprisedofaggregatesofpairedhelicallamentsPHF,alpha-helicesofhyperphosphorylatedmicrotubulartauproteinAlonsoetal.,2001;Canevarietal.,2004;Sobowetal.,2004.Tau,whichnormallybindstubulin,isinvolvedinmicrotubuleformationandstabilizationwithinneuronsMudherandLovestone,2002.Followingphosphorylationbyproteinkinasescdk5orGSK3beta,taudissociatesfrommicrotubulesandformsPHFsMaccionietal.,2001.Finally,accumulatingPHFsdestabilizeintraneuronalmicrotubulesandeventuallyreplacethemicrotubuleswithpathogenictanglesMudherandLovestone,2002.1.4DiagnosisofAlzheimer'sDiseaseDementia{theprogressivedeteriorationofcognitivefunction{mayresultdi-rectlyfromprimarydiseasesofthebrain,orasaconsequenceofotherdiseasestates.Themostcommontypesofdementia,distinguishedbyageofonsetandspecicpatternofstructuralandfunctionalpathology,include:Alzheimer'sdisease,fron-totemporaldementia,HIV-associateddementia,Lewybodydementia,andvasculardementia.DementiaisoftencomorbidwithHuntington'sdisease,Parkinson'sdis-ease,progressivesupranuclearpalsy,neurosyphilis,andseveralpriondisorderse.g.,Creutzfeld-Jakobdisease,Gerstmann-Straussler-Scheinkersyndrome.Inaddition,certainenvironmentaltoxinse.g.,lead,industrialsolvents,metabolicdisorderse.g.,hypothyroidism,vitaminB12deciencyandbraininjuriese.g.,subduralhematoma,normal-pressurehydrocephaluscanproducedementiawhich,insomecases,maybereversiblewithtreatment.Severalpsychiatricconditions,including6

PAGE 17

depression,mayresembledementiaaswell.DierentialdiagnosisofAlzheimer'sdis-ease,therefore,requiresacompletepersonalandfamilymedicalhistory,andconsider-ationofcoexistingmedical/psychiatricconditionswhichsharesymptomsincommonwithAlzheimer'sdisease.Thecompletediagnosticprocessinvolvesseveralstagesinincreasingorderofinvasiveness.Generalcognitiveimpairmentappearsgraduallyovertime,andprobableAlzhei-mer'sdiseaseisdiagnosedthroughbehavioralassessment,typicallyasclinicalob-servationandinterview-basedexamination,overaperiodofweeksormonths.TheovertsymptomsofAlzheimer'sdiseasearecalledtheFourAs"{agnosiathein-abilitytorecognizeobjectsorfamiliarpeople,amnesiatheinabilitytoremember,aphasialinguisticincompetence;theinabilitytounderstandorcommunicateusinglanguage,andapraxiatheinabilitytoperformactivities.Physicalandpsycho-logicalexaminationsarenecessaryfordierentialdiagnosis,toruleoutalternativedisorderswithsimilarpresentations,suchasclinicaldepression.Interviewswiththepatient,familymembers,andcaregiversalsoprovidediagnosticinformation.Inaddition,electroencephalographicandbrainimagingtechniquesareusedtosupple-mentdiagnosis.Currently,thedenitivediagnosisofAlzheimer'sdiseaserequiresmicroscopicexaminationforthecharacteristiclesionsatautopsy.1.4.1PsychometricAssessmentEvaluationofprobableAlzheimer'sdiseaseinbothclinicalandcommunity-livingsettingsrequiresstandardizedpsychometrictestinginstruments.Cognitiveassess-mentinventoriescommonlyusedinclinicalsettingsincludetheMini-MentalStateExaminationMMSE;Folsteinetal.,1975andtheBlessedOrientation-Memory-ConcentrationtestBOMC;Katzmanetal.,1983.TheMMSEconsistsofathirty-pointquestionnaireofitemsrepresentingvecognitivecomponents:timeandplace7

PAGE 18

orientation,registrationabilitytoidentifyandrememberthreecommonobjects,attentionandmentalcalculatione.g.,serialsubtraction,short-termmemoryrecallofthethreeobjectspresentedearlier,andlanguageabilitye.g.,repeatphrase,followdirectionsgivenbyexaminer.ScoresarecomparedagainstnormsestablishedforageandeducationallevelCrumetal.,1993.TheBOMCcontainssixitems,ad-dressingthreecognitivedomains:timeorientation,concentrationanalogoustotheMMSEattentionandmentalcalculationcomponent,andmemoryrecall.Thesim-ilarcontentoftheseinstrumentsisreectedinthestrongintercorrelationbetweenexamineescoresFillenbaumetal.,1987;Zillmeretal.,1990.High-throughputscreeningtechniques,abbreviatedversionsofmentaltestbatteries,areunderdevel-opmentforecient,large-scaleevaluationofprospectivepatientgroups.Mnemonicmeasures,forinstance,areusefulforAD-screeningaswellasindierentialdiagnosis.Forexample,ADpatientsscorelowerthanMCIpatientswho,inturn,scorelowerthanage-matcheddepressedpatientsonvisualassociationcued-recalltestsDierckxetal.,2007,usingdiscriminantanalysis5%overallaccuracy.Semanticuencytaskse.g.,free-responsenamingofexemplarswithinatargetcategoryareparticu-larlysensitivediscriminatorsbetweennormalagingandAlzheimer-associateddecitsCerhanetal.,2002;Salmonetal.,2002.Forexample,intestsofverbaluencynamesofanimals,wordshavingsamerst-letter,normal-agedindividualsgeneratemorewordsandproducelargerclusterswithincategories,relativetomild-ADpa-tientsGomezandWhite,2006.Inaddition,comprehensivecognitiveassessmentprotocolshavebeenestablishede.g.,CERAD;Morrisetal.,1993forclinicaleval-uation,therapeuticmonitoring,andepidemiologicalstudiesofADpatients.TheConsortiumtoEstablishaRegistryforAlzheimer'sDiseaseCERADhasdevel-opedamultimetricneuropsychologicalbatteryfordiagnosingprobableAlzheimer'sdisease,aswellasforcharacterizingtheseverityofcognitiveimpairmentMorris8

PAGE 19

etal.,1993;StraussandFritsch,2004,consistingoftheMMSEandsixadditionalperformancemeasures:modiedBostonNamingTest,verbaluencytest,construc-tionalpraxisdesigncopying,andthreeword-listlearningmeasuresimmediaterecall,delayedrecall,andwordrecognition.FactoranalyticstudiesoftheCERADinventoryStraussandFritsch,2004;JonesandAyers,2006suggestasinglefactorrepresentingoverallcognitiveperformanceinprobable-Alzheimer'ssubjects,howeveratwo-factorsolutionwasobtainedwhenage-matched,non-ADsubjectsareincludedintheanalysisJonesandAyers,2006.Indeed,researchershaveextractedtwo-e.g.,Jacobsetal.,1994,three-e.g.,Kanneetal.,1998,andevenve-factore.g.,O'Donnelletal.,1988solutionsusingmultiplepsychometricinstruments{includingtheMMSEFolsteinetal.,1975,SPMSQPfeier,1975,andACADSevushetal.,1991;theapparentmultidimensionalityofcognitioninAlzheimer'sdiseasehasbeeninterpretedintermsofeitherpremorbidheterogeneityorAD-relatedheterogene-ityFisheretal.,1999;Sevushetal.,2003.Twohundredsixty-onepatientswithprobable-ADcompletedtheMMSE,SPMSQ,andACADexaminationsSevushetal.,2003;subsequentfactoranalysisprincipalcomponents,Varimaxrotationiden-tiedtwogeneral-cognitivefactors,relatedtoAlzheimer'sprogression,andathirdfactorreectingpremorbidcharacteristicscorrelatedwithdemographicvariables.Inaddition,Alzheimer'sdiseasehasbeenconsideredaheterogeneousdisorder,hav-ingdiverseovertbehavioralandcognitivemanifestationsCummings,2000.ThecognitiveimpairmentassociatedwithAlzheimer'sdiseasetendstoincreaseinbothseverityandscope,mirroringtheinsidiousprogressionoftheunderlyingneuropathol-ogyMorrisetal.,1989;Cummings,2000.Incontrast,variouspsychologicalandbehavioralsymptomsofAlzheimer'spathologymayappearduringdierentstagesJostandGrossberg,1996.ThemultidimensionalcharacterofAlzheimer'sdisease,forinstance,wassuggestedbyafactoranalysisprincipalcomponentofeighteen9

PAGE 20

standardbehavioral/cognitivemeasuresinanADoutpatientDSM-IV,NINCDS-ADRDAcriteriasampleN=244;Spallettaetal.,2004,whichidentiedsevendistinctfactorsandtheirassociatedbehavioralvs.cognitivedimension:generalcognitivecognitive,hyperactivitybehavioral,psychosisbehavioral,constructiveabilitycognitive,anxietybehavioral,mood-excitementbehavioral,andmood-depression/apathybehavioral.Theseresults,collectively,suggestthatbehavioralandcognitivecomponentsrepresentindependentdimensionsofADSpallettaetal.,2004,basedonmultipleconventionalneuropsychologicalassessmenttechniques,andunderscoretheprimacyofasingleunderlyingglobal"constructofgeneralcognition.Finally,althoughbothpsychometrice.g.,MMSEandbehavior-basede.g.,DAFSevaluationprotocolshavedemonstratedeectivenessfordierentialdiagnosisandtherapeuticassessmentinAlzheimer'sdisease,interpretingthecomplexinterplaybetweencognitiveandbehavioralprocesses{asrevealedthroughfactoranalyses{remainsacontroversialareaofresearche.g.,Ownbyetal.,2004.CooperativeeortstostandardizeclinicaldiagnosisofAlzheimer'sdisease,bytheNationalInstituteofNeurologicalandCommunicativeDisordersandStrokeNINCDSandtheAlzheimer'sDiseaseandRelatedDisordersAssociationADRDA,ledtothedevelopmentofasetofcriteriabasedonclinicalobservation,neuropsycho-logicaltesting,andhistopathologicevidenceMcKhannetal.,1984;Blackeretal.,1994.Thepresenceofadementiasyndrome,inadditiontoprogressivecognitiveim-pairment,occurringintheabsenceofotherdiseasescapableofproducingdementia,isrequiredforadiagnosisofprobableAlzheimer'sdisease;post-mortemmicroscopicexaminationofbraintissueisrequiredtoconrmdeniteAlzheimer'sdisease.Re-centadvancesindiagnosticimagingtechniques,however,havepromptedtherevisionoftheNINCDS-ADRDAcriteria,andanewproposalisunderreviewDuboisetal.,2007.Finally,neuropsychologicaltestsaremodestpredictorsofdailyfunctionalsta-10

PAGE 21

tusinAlzheimer'spatients.InastudyFariasetal.,2003involving42outpatientsdiagnosedwithpossible-orprobable-AD,cognitiveperformanceimmediate-anddelayed-recall,attention,visuospatial,executivefunctioning,andpraxismeasureswassignicantlycorrelatedwithfunctionalstatusDAFSscale;Loewensteinetal.,1989.TheDAFSscaleLoewensteinetal.,1989,1995isahighly-reliablebothinter-raterandtest-retestbehavior-basedevaluationofdailylivingskillsinfamiliardomains,suchas:tellingtimereadingaclock,handlingmoneywritingacheck,balancingacheckbook,makingchange,andhouseholdroutinesaddressingandmailingaletter,usingatelephone.TheDAFSisausefulscreeningdevicefordif-ferentialdiagnosisamongnormal-aged,depressed,andprobable-ADindividuals;ADpatientsscoresignicantlylowerthaneithernormal-agedordepressedindividualsonallfunctionalmeasuresexcepttellingtimeLoewensteinetal.,1989.1.4.2BiomarkersAdvancesinbioinformaticsmethodologyenableinvestigatorstoexaminegene-levelgenomicandprotein-levelproteomicfeaturesofnormalandpathologicalphenotypes,includingneurologicalsyndromesMirnicsandPevsner,2004.Manip-ulationsofgeneticexpressionintransgenicanimals,forinstance,whichalterneuronalstructureand/ormolecularsignallingpathwaysaremanifestedthroughdistinctcog-nitivegenotypesFlint,1999.Candidate-geneassociationstudiesandgenome-widescanssuggestthathumanmemoryisapolygeniccognitivetraitdeQuervainetal.,2003;Eganetal.,2003;Papassotiropoulosetal.,2005;deQuervainandPapas-sotiropoulos,2006,withheritabilityestimatedat50%McClearnetal.,1997.Forexample,agenome-widescanPapassotiropoulosetal.,2006involvingover500,000single-nucleotidepolymorphismsinpooledDNAfrom341youngadults,stratiedbyperformanceinaverbaldelayed-recalltask,linkedbetterperformanceafterboth11

PAGE 22

5-mindelayand24-hrdelaywiththepresenceoftheKIBRArs17070145Tallele;anindependentsampleof256adultsassociatedthisallelewithsuperiorperformanceonstandardtestsofepisodicmemoryBuschke'sSelectiveRemindingTest,ReyAuditoryVerbalLearningTest,althoughnosignicantallele-dependentdierenceswereobservedintheWisconsinCardSortingTestorthePacedAuditorySerialAttentionTaskPapassotiropoulosetal.,2006.Hence,asinglegeneticmutationKIBRArs17070145isassociatedwithindividualdierencesinepisodicmemory,butnotattention,executivefunction,orworkingmemoryPapassotiropoulosetal.,2006.Furthermore,gene-associateddierencesinbrainfunctione.g.,memory,intelligencearepartlydeterminedbygene-inuencedstructuraldierencesinthebrainTogaandThompson,2005;Zimmer,2008.Additionally,geneticevidencefromanimalstudiesunderscorestheassociationbetweengenesandcognitiveability.Geneticdierencesinspatiallearningabilityinmicehasbeenexaminedusingmul-titraitanalysisSteinbergeretal.,2003inDBA/2andC57BL6/Jmousestrains,andrevealstwoquantitativetraitloci,onchromosomes4and12,areassociatedwithMorriswatermazeretentiontimespentandnumberofannularcrossingsinformerplatform-containingquadrant,aswellasspatiallearningrateswimpathdistance,latencytoreachplatformlocation,measuredintheprobetrialSteinbergeretal.,2003.Techniquesbasedondatamininge.g.,Bayesianstatisticalmodelshavebeenappliedtohumanbrain-specicgenenetworkdatabasescompiledfromgenomicandproteomicstudiesinordertoidentifycandidategenesforneurologicaldisordersLiuetal.,2006.Genomescreeninginknown-ADfamilies,forexample,hasidentiedmutationsinthreegenesassociatedwiththeearly-onsetfamilialADAPP,PSEN1,andPSEN2;e.g.,TanziandBertram,2001andalate-onsetAD-associatedpoly-morphismofApoEApoE-4;chromosome19q13.Inaddition,linkageanalysis12

PAGE 23

suggestsasmanyastwelvegeneticlocicontributetoADq23,3p26,4q32,5p14,6p21,6q27,9q22,10q24,11q25,14q22,15q26,and21q22;Blackeretal.,2003.Mi-croarrayanalysisofplasmasamplesfromnormal-controlandAD-individualsidenti-edeighteenproteinswhichdiersignicantlyinconcentrationi.e.,ANG-2,CCL5,CCL7,CCL15,CCL18,CXCL8,EGF,G-CSF,GDNF,ICAM-1,IGFBP-6,IL-1,IL-3,IL-11,M-CSF,PDGF-BB,TNF-,andTRAIL-R4,anddistinguishbetweenthegroupswithapproximately90%accuracyRayetal.,2007.Thesesignalingproteinsareassociatedthroughtwoindependentregulatorypathways:onesetisrelatedbyTNF-tumornecrosisfactorandM-CSFmonocyte-colonystimulatingfactor,whiletheothersetcentersaroundEGFepidermalgrowthfactorRayetal.,2007.Theinvolvementoftheseproteinsinimmunoreactivity,hematopoiesis,andapoptosisisconsistentwithrecentndingsshowingincreasedhematopoieticcellactivityintheAlzheimerbrainbothhumanADandinAlzheimer'stransgenicmice;Wyss-Coray,2006;Simardetal.,2006;BritschgiandWyss-Coray,2007,aswellasAlzheimer-associatedabnormalitiesinapoptoticpathwaysLeBlanc,2005.Thediagnosticutilityofmicroarrayanalysisisunderscoredbypostmortemfrontalcor-ticalRNAanalysesinnormal-agedandADpatientsWalkeretal.,2004.Adatamining-basedclassierBioMinerTMconrmedtheexpressionof17neuropathology-associatedgenesDTNA,B2M,APLP1,C4B,LIMS2,IGHM,GRP58,KRT8,ATF4,RANGAP,GSTM2,TU3A,ADD3,FTL,HBB,CLU,andHBG2incorticalsamplesfromADpatientsWalkeretal.,2004.Similarly,geneprolingcDNAmicroarrayof18-month-oldAlzheimer'strans-genicmiceAPPswandage-matchedcontrolanimalshasidentied52dierentiallyexpressedgenesJeeetal.,2005.Subsequentexaminationofage-relatedpatternsingeneticup-anddown-regulation,comparingaged18month-oldandyoungmonth-oldtransgenicandnormalmice,revealed48and40dierentiallyexpressed13

PAGE 24

genesbetweenagedandyounganimalsofeachgroup,respectivelyJeeetal.,2007.Age-associatedchangesinthegeneticprolesareconsistentwithinammatoryre-sponse,increasedoxidativestress,anddecreasedneurotrophicsupportLeeetal.,2000.Dierential2Delectrophoresisofwhole-corticaltissuefrom14-month-oldAlzheimer'stransgenicmiceAPP/PS1identiedsignicantdierencesinthecon-centrationsof15proteins,relativetowild-typeanimalsSizovaetal.,2007.Inadditiontointraneuronalandmembraneproteinsassociatedwithsynapticfunctionandaxonalgrowth,proteinsinvolvedinglialresponseandinammation,cholesterolmetabolism,andoxidativeresponsewereincludedSizovaetal.,2007.1.4.3NeurologicalBasisofMemorySystemsAnatomicalcomponentsofthemammaliannervoussystemformacomplex,in-terconnectednetwork.Understandingneuropathologies,suchasAlzheimer'sdisease,requiresthoroughstudyoftherelationshipbetweenstructureandfunctionatdif-ferentscales.Indeed,memory{thestorageandretrievaloflearnedinformation{hasbeenexaminedandmodeledinhumansandotheranimalsondierentscales,withinbothnormalandpathologicalstates,andinavarietyofcontexts.Onacel-lularscale,forexample,simultaneousstimulationoftwoneuronsenhancessynaptictransmissionbetweentheneuronsthroughacomplexcascadeofintracellularproteintranscriptionalandmolecularsignallingeventsKandel,1979;Sweatt,1999;Kandel,2001,consequentlyimprovinglong-terminterneuronalcommunicationlong-termpotentiation,"LTP;Lynch,2004.Direct-stimulationstudies,forexample,showthatshort-termmemoryformationisdependentuponLTPinthehippocampuse.g.,BlissandLomo,1973,andpharmacologicagentswhichinterferewithhippocam-palLTPe.g.,byblockingNMDAreceptorsinratsalsoimpairspatialmemoryperformanceMorriswatermazeMorrisetal.,1986.Inaddition,micereceiv-14

PAGE 25

ingmedialfrontalcorticalinjectionsofanisomycinaproteinsynthesisinhibitorexhibitspatialmemoryacquisitiondecitsradialmazenonmatching-to-placetaskTouzanietal.,2007,suggestingtheroleofproteinsynthesisduringlearning.Amacroscale,informationprocessingmodelofmemoryusesstoragedurationasthedistinguishingcriterionforatripartiteorganizationofmemorysystems,consistingof:sensorymemoryregistryofambientenvironmentalstimuli,typicallyavailableforaverybriefduration;Sperling,1960;shorttermmemorylimited-capacitytem-porarystore,BaddeleyandHitch'sworkingmemory,"1974;Miller,1956;Brown,1958;PetersonandPeterson,1959;and,longtermmemoryhigh-capacity,per-manent"referencememory;Landauer,1986.Longtermmemoryissubsequentlydividedintodeclarativeexplicitmemoryandnondeclarativeimplicitmemorycomponents,eachhavingdistinctneuroanatomicalsubstratese.g.,Schacter,1987;ThompsonandKim,1996;SquireandZola,1997;Miyashita,2004.Indeed,muchofwhatisknownaboutthestructureofhumanmemorysystemsisinferredfromclinicalstudiesofamnesicse.g.,ShalliceandWarrington,1970;Cohenetal.,1985;GrafandSchacter,1987;Perianietal.,1993,whoexhibitdissociationsyndromesamongcom-ponentsystems.Theinabilitytoformnewlong-termmemoriesthroughexperienceanterogradeamnesia,forexample,isassociatedwithhippocampallesionsortempo-rallobedamageeitherdeliberatelybysurgery,oraccidentallythroughinjurye.g.,KandelandPittenger,1999.Declarativememoryinvolvesmedialtemporallobe,hippocampal,andprefrontalcorticalstructuresKandelandPittenger,1999,andconsistsoffactualinformationsemanticmemoryandautobiographical,experience-basedknowledgeepisodicmemory;thesemantic-episodicdistinctionwassuggestedbyTulving972,1985.Bycontrast,nondeclarativememoryinvolvesthecerebel-lumandstriatum,andencompassesprocedural,skill-basedknowledge,includingconditioning;associativeprimingMeyerandSchvaneveldt,1971,articialgram-15

PAGE 26

marlearningcorrectidenticationofrule-consistentstringsoflettersformedfromanite-stategrammar;Reber,1967,categorylearning,andsequencelearninge.g.,serialreactiontimetaskareexamplesofimplicitmemoryphenomena.Withinacomprehensiveassessmentoflearningandmemoryfunction,domain-specicmemorytasksprovideevidenceforthedierentialeectsofnormalandpathologicali.e.,ADagingonhumanmemorysystems.Forexample,comprehen-sivememorytestsadministeredtoearly-to-moderateADpatientsandage-matchednormalindividualsStopfordetal.,2007identiedvedistinctmemorydomainspersonalmemory,recognition,verbalrecall,visualrecall,andworkingmemoryinwhichimpairmentwithinonedomainmayoccurindependentlyoftheotherdomains.Inacomparisonofdierentialimpairmentacrossmemorysystems,Mitchell89examinedtheeectofnormalagingintwoadultcohortscross-sectionalstudy;19-32y/oyoung"and63-80y/oolder";N=48,eachusingmultipleperformancemeasures:episodicmemoryfreerecall,recognition,intrusionsduringrecall,andlongitudinalchangeinrecall;Underwoodetal.,1978;proceduralmemoryrepeti-tionpriminglatency;GrafandSchacter,1985;and,semanticmemoryvocabulary,picture-naminglatencyanderrors.Althoughnosignicantage-relateddierenceswereobservedforeitherproceduralorsemanticmemory,youngeradultsperformedbetterinrecallandrecognitionmeasuresofepisodicmemoryMitchell,1989;addi-tionally,factoranalysisofthemultimetricassessmentconrmedthetripartitemem-orytaxonomy.Episodicmemoryfreerecall,wordrecognitiondecitshavebeenobservedinpreclinicalADpatientsBackmanetal.,2001,uptosixyearspriortodiagnosis,aswellasdecitsinperceptualspeedandexecutivefunctioningBack-manetal.,2005.Noshort-termmemorydigitspanimpairmentswerefoundinpreclinicalADcases,relativetoage-matchednormalindividuals,howeverBackmanetal.,2001.16

PAGE 27

1.4.4ElectroencephalographyAlandmarkdiscoveryinneurology,theconstructionofaminimally-invasiveelec-tricalinterfacebetweenthelivingnervoussystemandarecordingapparatus,enabledinvestigatorstomonitorensembleneuralactivityinreal-time.Electroencephalog-raphyEEG,rstdescribedbyHansBergerin1924,involvesrecordingelectricalpotentialsfromanarrayofelectrodesdistributedacrossthescalpsurfacefollowingastandardizedgridpatternRowanandTolunsky,2003.Thepotentialsrecordedfromeachelectrodereectthecollectiveelectricalactivityi.e.,APs,EPSPsandIP-SPsoftheunderlyinglocalpopulationofneurons,toadepthofseveralcentimeters.TheEEG,therefore,representsregionalbrainelectricalpotentiale.g.,microvoltsasafunctionoftimee.g.,milliseconds.FourcharacteristicsignalpatternshavebeendescribedinthehumanEEG,somewhatarbitrarilybasedonfrequencyranges,whichcorrespondtodierentmentalstates.Betaactivityfast-wave",indicativeoffocusedmentalactivityandattention,hasrelativelyhighfrequencygreaterthan12Hzandlowamplitude.Alphaactivity,recordedinrelaxed,awakesubjectswiththeireyesclosed,exhibitsafrequencyof8to12Hzwithamplitudebetween10and50microvolts.Theprominenceofthealpharhythmintheoccipitalregion,anditsresponsivenesstoopeningandclosingoftheeyes,suggestsanassociationwithvisualprocessingAdrian,1935.Thetawavesofbetween4and7Hz,anddeltaslow-wave"activity,oflessthan4Hzfrequency,aretypicalofdrowsinessandsleep.VisualinspectionistheconventionalmethodforinterpretingEEGsrecordedfromsubjects.Thepresenceorabsenceofspecicfeaturesorpatternsofactiv-ity,asdepictedonagraphicalrecordingtrace,representsneurologicaldiagnosticcriteria.HumanEEGsrecordedduringabroadrangeofbehavioralstates{bothnormalandpathological{havebeenexaminedinthismanner.Forexample,in-17

PAGE 28

creasedfrontalmidlinethetaactivityisassociatedwithworking-memoryintensetasksGevinsetal.,1997,1998.Sleep-staging,forinstance,traditionallyreliesonvisualrecognitionbyanexpertofuniquewaveformsandsustainedperiodsofslow-waveactivityintheEEGtracee.g.,RechtschaenandKales,1968.Analternative,computationalapproach,nonlineardynamicalanalysisNDA,utilizesmathematicaltechniquesadaptedfromengineeringforstudyingcomplex,time-basedphysicalphe-nomenaPradhanandDutt,1993;PritchardandDuke,1995.Usingthisapproach,thedynamicsoftheEEGsignalarecharacterizedmathematicallyintermsofdi-mensionalcomplexity"representedasthecorrelationdimension,D2;GrassbergerandProcaccia,1983.Dimensionalcomplexityisasummarystatistic,capturingtheoverallcharacterofcorticalactivityincl.nonlinearcommunicationbetweenindi-vidualneurons,aswellasbetweenneuronpopulationsinasinglemeasure.Itisnoteworthythatthepresenceofdynamicallychaoticactivitythroughoutthebrainmaybeessentialfor{andindicativeof{normalfunctioning,includingconsciousnessFelletal.,2003.ThespatiotemporalcomplexityoftheEEGmirrorstheintricateneurophysiologicalsubstrateforbehavioralandcognitiveprocessing.Indeed,withinthenervoussystem,synchronyandcoherenceareoftenassociatedwithpathologye.g.,epileptiformspikingandseizuresSkardaandFreeman,1987;Doyon,1992.Inadditiontoprovidingtheoreticalinsightsonspatiotemporalphenomenainthebrain,dimensionalcomplexitymeasurescanbeusedtodescribecorticalstatesforidentifyingneuralcorrelatesofbehaviorandcognitiveactivity,toexaminetherapeuticecacyandexperimentaltreatmenteects,andtoassistinthediag-nosisofneuropathology.Forexample,levelsofconsciousnesscanbereliablydis-tinguishedthroughNDAtechniques,facilitatingcomputer-assistedautomatedEEGmonitoringandrecording.Inhumans,assleepproceedsfromNREMnon-REM"Stage1throughNREMStage4,amonotonicdecreaseinmeasureddimensional18

PAGE 29

complexityD2hasbeenobservedPradhanandSadasivan,1996.Inaddition,D2measuresduringREMareslightlyhigherthanD2duringwaking.AutomateddiscriminationamongNREMstagesusingD2measuresfromsampledEEGdataiscomparableorsuperiortoexpert-stagedhypnogramsPradhanandSadasivan,1996.LowerD2measures,relativetorestingconditions,areobservedduringmedi-tativestatesAftanasandGolocheikine,2002particularlyoverthemidlinefrontalandcentralregions.Sleep-deprivedsubjectsalsoexhibitlowerD2,whichmayrepre-sentsub-optimalinformationprocessingcapacityJeongetal.,2001b.Inaddition,task-dependentchangesinD2mayberelatedtoindividualintellectualabilitye.g.,JausovecandJausovec,2000;Lutzenbergeretal.,1992;Molleetal.,1999,consis-tentwitharesource-demandmodeloffunctionalprocessinginthebrain.NLDA-basedstudiesofneuropathologies,includingAlzheimer'sdisease,suggestthatdi-mensionalcomplexitymaybeamoresensitivediagnosticmarkerthantraditionalexpert-basedEEGinterpretivemethodse.g.,GallezandBabloyantz,1991;Jeongetal.,1998;Jeong,2004.DecreasedD2wasfoundinclinicalpatientswithposttrau-maticstressdisorderPTSD,suggestiveofimpairedcorticalinformationprocessingChaeetal.,2004.PatientsdiagnosedwithschizophreniadonotexhibitinitialtransientD2changes,comparedwithnon-schizophrenicindividuals,whenperform-ingacognitively-demandingtaskKirschetal.,2000,suggestingimpairedabilitytorespondadaptivelytochangingcognitivedemands.TheprimaryEEGndingsinADpatientsinclude:increaseddiuseslowactivityandslowingofthedominantposteriorrhythmBrenneretal.,1988;Jeong,2004,decreasedalphaandbetaactivityLetemendiaandPampiglione,1958;Jeong,2004,increasedthetaanddeltaactivityBrenneretal.,1986;GiaquintoandNolfe,1986,anddecreasedsynchronizationcoherenceinthealphaandbetabandsDunkinetal.,1994;Koenigetal.,2005.Theseabnormalitiescorrelatewithdiseaseseverity19

PAGE 30

Hughesetal.,1989;Kowalskietal.,2001,andreectconcomitantstructuralandfunctionalimpairmentofthecerebralcortex.UnderscoringcholinergicinvolvementinAlzheimersdiseaseisthendingthatyoung,healthysubjectsreceivingthemus-cariniccholinergicantagonistscopolaminealsoexhibiteither/bothincreasedslow-waveanddecreasedfast-waveactivityEbertandKirch,1998;Ebertetal.,2001andtransitoryAlzheimer-likememoryimpairmentWesnesetal.,1988.Inaddition,ADpatientsreceivingthecholinergicagonistnicotineshowincreasedfast-waveactivityanddecreasedslow-waveactivityKnottetal.,2000,andmild-ADpatientsreceiv-inglong-termtreatmentwithdonepezilshowdecreasedthetaactivityKoganetal.,2001.NLDA-basedstudies,usingsingle-channelaswellasmulti-channelEEG,showdecreasedD2inAD,suggestingreducednonlinearcellularcommunicationorcouplingbetweencorticalregionsJellesetal.,1999;Jeongetal.,2001a.1.4.5DiagnosticMedicalImagingMedicalimagingtechniquesprovideusefuldiagnosticinformation.Thesemeth-odsinclude:computedtomographyCT,magneticresonanceimagingMRI,func-tionalmagneticresonanceimagingfMRI,positronemissiontomographyPET,andsinglephotonemissioncomputedtomographySPECT.Forexample,anatom-icalstructuralchangesinthebrainaredetectedusingCTandMRI,whilephysio-logicalmetaboliccorrelatesofbrainfunctionarestudiedusingPETandSPECT.RegiospecicatrophyofthehippocampusandentorhinalcortexhavebeenidentiedinlongitudinalstudiesofAlzheimerspatientsusingCTandMRIJacketal.,1997;Xuetal.,2000;Barnesetal.,2004;Pennanenetal.2004underscoringtheutilityofthesemethodsformonitoringtheprogressofADinpatientsovertime.Localizeddecreasesinglucosemetabolism,measuredthroughPETstudies,canidentifyareascompromisedbyAlzheimersdiseaseMinoshima,2003.ModiedPETtechniques,20

PAGE 31

usingaradiolabeledligandwhichbindstobetaamyloid,havebeenusedtolocalizeamyloidplaquesKlunketal.,2003;Klunketal.,2004.CorrelationbetweeninvivoandinvitroPETimagingstudiesusingtheradiotracerPittsburghCompound-BIkonomovicetal.,2008,forinstance,showhighlyselectivebindingforinsolublebrillarbetaamyloiddeposits.AsimilarapproachisbeingdevelopedforPETimagingofneurobrillarytanglesMathisetal.,2004.Additionally,arenementofSPECTHMPAO-99TcSPECT;DeFigueiredoetal.,1995hasdemonstratedutilityfordierentialdiagnosisofADfrombothvascular-andfronto-temporaldementiasDougalletal.,2004,with71.3%5.9%and71.5%.2%sensitivitiesandspecicities,respectively.Earlyperceptualandinformationprocessingevents,andtheirneuralsubstrates,areexaminedthroughdiagnosticmedicalimagingtechniques.Moreover,diagnosticimagingduringtaskperformancehelpsrevealtheneuroanatomicalsubstrates,con-nectivity,andmechanismswhichunderliecomplexcognitivebehavior.Forexample,theStroopcolorwordinterferencetask,inwhichsubjectsidentifycongruencebe-tweencolor-wordi.e.,nameofacolorandword-colori.e.,colorofaprintedwordstimulipresentedvisuallyStroop,1935,requirescoordinatedvisualandverbalpro-cessing;ADindividualsexhibitdecitsinthistask,relativetoage-matchednormalsubjectsBondietal.,2002;Amievaetal.,2004.TheneuralcircuitryinvolvedintheStroopcolorwordinterferencetaskhasbeenexploredusingfMRIPetersonetal.,1999;Leungetal.,2000.RegiospecicactivationpatternswithinthefrontalandanteriorcingulatecortexfollowatemporalsequencecorrelatedwithvisuospatialandverbalprocessingofthetargetstimuliLeungetal.,2000.LeftprefrontalcorticalactivationwasidentiedduringverbaluencytasksusingfMRIinnormalindividu-alsSchlosseretal.,1998,aswell.Additionally,neuroimagingcanrevealparallelsbetweennascentbehavioralfunctionalimpairmentandneuropathology,andsug-21

PAGE 32

gestpatternsoffunctionalcompensationduringearlyADe.g.,Posneretal.,1997.Forexample,age-matchednormalandearly-ADindividualsperformedanepisodicmemorytaskwhileundergoingfunctionalmagneticresonancefMRIimagingGronandRiepe,2004;inagradedprogression,increasingfunctionalimpairmentpoorercognitiveperformancewasassociatedwithdecreasingposteromedio-temporalac-tivation,acrossnormalandADsubjects.Intactoperationalcomponentsofwork-ingmemory{executivecontrolinformationencodingandretrievalprocessesandactivemaintenancepreservingimmediateavailabilityofinformation{areneces-saryforshort-termstorageandmanipulationofinformatione.g.,BaddeleyandHitch,1974;Baddeley,1992.FunctionalMRIstudiesinhumansperformingwork-ingmemorytasksindicatebothfrontalandparietalcorticalinvolvementintheactivemaintenancecomponentCohenetal.,1997.PositronemissiontomographyPETstudiesofhumansubjectsduringepisodicmemorytasksyes-norecognitionrevealincreasedactivitywithintherightprefrontalcortexandanteriorcingulateNybergetal.,2000.Inaddition,PETduringanepisodicmemorytaskinvolvingnormal-agedandmildlycognitively-impairedMCIpatientsshoweddierentpatternsofregionalactivationbetweenthetwogroupsMoulinetal.,2007;thetaskconsistedoftwosuccessiveencodingtrialssemantically-relatedpairsofwordsfollowedbyaretrievaltrial.Althoughbothgroupsexhibitedleftfrontallobeactivationdur-inginitialencoding,onlythenormal-agedindividualsshowedactivationduringthesecondaryencoding.Inaddition,MCIindividualsshowedactivationinthevisualcortexduringsecondaryencoding,whilenormal-agedindividualsdidnot.Duringretrieval,MCIindividualsdidnotexhibitrightfrontalorlefttemporalactivationincontrasttonormal-agedindividuals,althoughtheyshowedmoreextensiveleftfrontalcorticalactivation,relativetonormal-agedindividualsMoulinetal.,2007.22

PAGE 33

AnimportantgoalofmedicalresearchinAlzheimer'sdiseaseisthedevelopmentofearlydiagnostictechniques,tooptimizetheecacyoftreatmentandintervention.MethodswhicharebothsensitivetoearlybehavioralandpathologicalindicationsandselectiveforAlzheimer'sdiseasearehighlydesirable.Contentanalysis,forin-stance,representsapotentialearlypredictor,providingameansforidentifyingtheprogressivedeclineoflinguisticabilitiesassociatedwithAlzheimer'sdiseaseForbesetal.,2004;Garrardetal.,2005;Vennerietal.,2005.Carefulexaminationofanauthor'swrittenworke.g.,thenalnovelofBritishwriterIrisMurdochcom-pletedpriortotheonsetofcharacteristicAlzheimersymptomscanreveallexicaldecitssuggestiveofincipientcognitiveimpairmentGarrardetal.,2005.However,whethersemanticandsyntacticcomponentsofverbalabilitydeclineseparatelyorinparallelremainsunclearGarrardetal.,2005.Alzheimer'sdisease,likemanyothernucleatingpathologiese.g.,priondiseases;Prusiner,1991;JarrettandLans-bury,1993;DeMageretal.,2002,exhibitsverysubtlemanifestationsinitsnascentform;prodromaldetectionofADiscomplicatedbecauseneuropathologyprecedesbehavioraldecits.1.5RiskFactorsforAlzheimer'sDiseaseAgingandfamilyhistoryhavebeenidentiedasmajorriskfactorsforAlzheimer'sdisease.However,Alzheimer'sdiseaseisnotaninevitableconsequenceofaging.Forexample,slightatrophyandsparseaccumulationsofhyperphosphorylatedtau,butnobeta-amyloidplaques,werefoundinthecortexandhippocampusofa115-year-oldwomanwhosecognitiveabilitiesremainedintactMMSEscorescomparableto60-75y/ountilherdeathfromcancerdenDunnenetal.,2008.AlthoughthecauseofAlzheimer'sdiseaseisunknown,itislikelyamultifactorialdisorderpossessingacombinationofdispositionale.g.,geneticandsituationale.g.,environmental23

PAGE 34

determinantse.g.,Mattsonetal.,2002.TwovariantsofAlzheimer'sdiseasehavebeenidentied:alate-onseti.e.,65yearsofage,oroldersporadic"form,whichencompassesthemajorityofcases,andanearly-onsetbetween30and65yearsofagefamilial"form,whichaccountsforapproximatelyvepercentofcases.Thesevariantssharesimilarbehavioralandpathologicalmanifestations,anddieronlyintheirageofonset.SeveralinalterablefactorscontributetothedevelopmentofAlzheimer'sdisease.ThelikelihoodofdevelopingAlzheimer'sdiseaseincreaseswithage.Femalesareathigherriskthanaremales.ThepresenceoftheApoE4alleleoftheApoEgeneincreasesriskinadose-dependentmanner,andlowerstheageofonsetofADVeurinketal.,2003;Yaoetal.,2004.Inaddition,theearly-onsetvariantofAlzheimer'sdiseaseisassociatedwithmutationsinseveralautosomal-dominantgenesSelkoeandPodlisny,2002;PardoandvanDuijn,2005;Chai,2007.LifeexperiencesandlifestylescaninuenceboththelikelihoodofdevelopingAlzheimer'sdiseaseaswellastheprogressofthedisorder.Ahistoryoftraumaticheadinjury,particularlyduringearlyadulthood,isassociatedwithincreasedriskFlemingeretal.,2003.Poorcardiovascularhealthhighbloodpressure,highcholesterolisalsoassociatedwithincreasedrisk.AlongitudinalstudyBuchmanetal.,2006examinedtherelationshipbetweenbodymassindexandseveralage-associateddementingneuropathologiesAD,cerebralinfarction,andLewybodydis-ease,andshowedasignicantassociationwithAlzheimer'spathologyevenaftercorrectingfordementia,chronicdiseases,andphysicalactivity.TherelativeriskofincidentdementiawasestimatedCoxproportionalhazardmodelinacohortstudySternetal.,1994;N=593non-dementedindividuals,aged60+y/o,inwhichin-creasedriskofdementiawasassociatedwithloweducationRR2.2,lowlifetimeoccupationalattainmentRR2.25,orbothRR2.87;thereducedriskofinci-24

PAGE 35

dentADassociatedwithincreasededucationaland/oroccupationalattainmentmayreecteitherthelimitationsofdiagnostictestingeducationinuencestestperfor-manceoranacquiredcognitivereservewhichdelaysormaskstheonsetofclinicaldementiaSternetal.,1994.Forreasonswhichremainunclear,highereducationalattainmentisalsoassociatedwithacceleratedcognitivedeclinewithadvancingageinADWilsonetal.,2004,particularlyintheexecutivespeedandmemorycogni-tivedomainsScarmeasetal.,2006a.Environmentaltoxinsmayalsoplayarole,particularlyduringearlydevelopment.Relativelyhighconcentrationsofcopper,forexample,havebeenfoundassociatedwithhumanamyloidplaquesandneuropilKowalik-Jankowskaetal.,2002.Additionally,aged3y/omonkeysexposedtoleadasinfantsexhibitgreaterexpressionofAPPandBACE1,aswellasbetaamyloiddepositsinthefrontalassociationcortexWuetal.,2008b;concomitantdecreaseinDNAmethyltransferaseactivityissuggestiveofanepigeneticimprintingmechanism.1.6RiskReductionStrategiesBenecialhealth-promotinglifestyles,includingincreasedlevelsofphysicalactiv-ityLarsonetal.,2006,mentalstimulatione.g.,reading,puzzle-solving;Vergheseetal.,2003,andsocialinteractionareassociatedwithdecreasedriskfordevelopingAlzheimer'sdisease,andmayhavelong-termprotectivebenetsFratiglionietal.,2004;KramerandErickson,2007.Elderlyaged65+yearsindividualswhoen-gageinregulari.e.,atleastthreetimesperweekphysicalexercisee.g.,walking,bicycling,swimmingaresignicantlylesslikelytodevelopdementia,comparedtoindividualswhoexerciselessfrequentlyLarsonetal.,2006.ParticipantsN=176;70+year-oldsinayear-longgroupinvestigativeprogramwhoengagedinmoderatedailyexercisereportedhighersubjectivemeasuresofquality-of-life,physicalwell-being,andphysicalself-perception,relativetomore-sedentaryparticipantsFoxet25

PAGE 36

al.,2007.Interestingly,althoughvoluntaryparticipationinleisurephysicalactivitiesisassociatedwithreducedADrisk,neitheroccupationalat-worknorcommutingto/fromworkplacephysicalactivityisassociatedwithreducedriskforeitherde-mentiaorADRovioetal.,2007.Ave-yearstudyWilsonetal.,2007involving700agedindividualsexaminedtherelationshipbetweendailycognitiveactivitylevelandADrisk,andreportedthatmorefrequentengagementincognitiveactivitywasassociatedwithreducedADincidence,andthatacognitively-inactivepersonwas2.6timesmorelikelytodevelopAD.Neuroprotectiveand/ortherapeuticeectsofdietarymodicationhavebeendemonstratedinbothlaboratoryusingAlzheimer'stransgenicanimalmodelsandclinicalhumanstudies.AMediterraneandiet,"lowinsaturatedfatandenrichedwithfruitsandvegetables,isassociatedwithreducedriskforAlzheimer'sdiseaseScarmeasetal.,2006b.Inaddition,specicdietarymicronutrientcomponentscaninuenceADrisk.TheincreasedintakeofvitaminsCandEisassociatedwithde-creasedADriskMasakietal.,2000;Morrisetal.,1998;theantioxidantpropertiesofthesevitaminshelpprotectneuronsfrommembranelipidperoxidation-induceddamageBermanandBrodaty,2004.Polyphenolcomponentse.g.,phenolicacid,avonoidsofphytochemicalextractsfromcertainspicesandfruitsincl.blueberriesandpomegranatesalsocontainantioxidantshavingbenecialeectsJosephetal.,1998;AggarwalandShishodia,2004;Josephetal.,2005;Hartmanetal.,2006.Dietsricherinsourcesofomega-3fattyacidse.g.,DHA,docosahexaenoicacid,suchassh,arealsoassociatedwithdecreasedADriskBourre,2004;Morrisetal.,2005.RegularintakeofalcoholandnicotinearealsoassociatedwithdecreasedADrisk,althoughthepotentialrisksfromchronicuseorabuseoftheseagentsisunderinvestigatione.g.,.Letenneuretal.,2004.26

PAGE 37

Regularconsumptionofcoee,tea,orothercaeine-containingproductsmayprotectagainstthedevelopmentofAlzheimer'sdiseaseandhelpmaintaincogni-tivefunctioningthroughoutlife.Caeineisanon-selectiveA1,A2Aadenosinereceptorantagonistwhichincreasesalertnessandarousale.g.,Nehligetal.,1992;Fredholmetal.,1999.InaPortugueseretrospectivestudy0-year,dailycaeineintakewassignicantlyassociatedwithreducedriskofADMaiaandDeMendonca,2002.ACanadianprospectivestudyinagedadultsshowedthatcoeeconsump-tionisassociatedwithreducedADriskLindsayetal.,2002.Across-sectionalstudyinagedadultsN=890female,meanage72.6yrs;N=638male,meanage73.3yrsexaminedtherelationbetweenmentalabilityonstandardizedtestsandself-reportedcoeeconsumption,andshowedasignicantassociationbetweenlifetimecoeeconsumptionandbettercognitiveperformanceinwomenJohnson-Kozlowetal.,2002.Nosignicantrelationbetweencoeeintakeandcognitiveperformancewasfoundinmen,norbetweendecaeinatedcoeeconsumptionandcognitiveperformancebyeithersexJohnson-Kozlowetal.,2002.Bycontrast,astudyVanGelderetal.,2007involving676Europeanmen,allbornbetween1900and1920,usedMMSEscoresinamixedlongitudinalmodeltoexploretheassoci-ationbetweencoeeconsumptionincups/dayandten-yearcognitivedecline.AsignicantdierenceinMMSE-scoredeclinewasobservedbetweencoee-consumersversusnon-consumers,withtheleastcognitivedeclineoccurringinindividualscon-sumingthreecupsperdayVanGelderetal.,2007.Theoptimalorally-administereddoseofcaeineinhealthyadulthumansforenhancingbothsensorimotor-cognitiveattentional,performanceandsubjectivemoodexperienceisapproximately250mgKaplanetal.,1997.StudiesinAlzheimer'stransgenicmiceAPPswstronglysupporttheputativeassociationbetweencaeineconsumptionandbothcognitive-protectionandAD-amelioration.Arendashetal.2006administeredcaeine.527

PAGE 38

mg/dayp.o.;equivalentto500mg/dayinhumantoAPPswmiceforsixmonths,beginningatfourmonthsofage.Caeine-treatedanimalsperformedsuperiortoage-matchedtransgeniccontrols,andcomparablytonon-transgenicanimals,acrossmultiplemeasuresofspatiallearning,workingandreferencememory,andrecogni-tion/identicationArendashetal.,2006.SignicantlydecreasedhippocampalAwasfoundincaeine-treatedtransgenicanimals,aswellArendashetal.,2006.Inaddition,althoughdietarycaeinesupplementationdidnotaecteithercorticalnorhippocampaladenosinereceptordensities,caeine-treatedanimalsexhibitednormalbrainadenosinelevels,whichmayexplaintheobservedcognitive-protectioneectArendashetal.,2006.Additionalstudiese.g.,Dall'Ignaetal.,2007suggesttheneuroprotectivemechanismofcaeineinvolvesblockadeofA2Areceptorsinthebrain.Chronicintakeofnon-steroidalanti-inammatorydrugsNSAIDs;e.g.,ibupro-fen,indomethacin,sulindacisassociatedwithadecreasedADriskMcGeeretal.,1996;Stewartetal.,1997;Zandietal.,2002.Theprimaryanti-inammatorymechanismoftheseagentsinvolvesinhibitionofthetwoisoformsofcyclooxygenaseCOX1andCOX2MarnettandKalgutkar,1999,however,NSAIDadministrationreducesA42levelsindependentlyofCOXinhibitionWeggenetal.,2001.StudiesinAlzheimerstransgenicmiceprovideamorecomplexportrait,however.Althoughibuprofen,urbiprofen,indomethacin,andsulindacreducebrainA42levelsinmice,manyotherNSAIDsincludingaspirin,naproxen,andketoprofendonotEriksenetal.,2003.Nevertheless,modulationoftheendogenousinammatoryresponsee.g.,bysuppressingmicroglialactivityremainsanactiveareaofpharmaceuticalresearche.g.,Kitzawaetal.,2004;McGeerandMcGeer,2007.28

PAGE 39

1.7GeneticsofAlzheimer'sDiseaseTheearly-onsetfamilial"variantofAlzheimer'sdiseaseresultsfrommutationswithinanyofthreeautosomal-dominantgenesPriceandSisodia,1998;Rocchietal.,2003;Brouwersetal.,2006:amyloidprecursorproteinAPP,chromosome21;Murrelletal.,1991;Mullanetal.,1992,presenilin-1PS1,chromosome14;Haasetal.,1999,andpresenilin-2PS2,chromosome1.Asdescribedearlier,thepolymorphicapolipoproteinEgeneApoE,chromosome19confersdierentialvulnerabilitytothemore-common,late-onset,sporadic"formofAD.Additionalgenes,identiedthroughmicroarrayanalysesofgenome-widescreens,arecurrentlyunderinvestigationtoelucidatetheirrolesinAlzheimer'sdisease.Inaddition,theassociationbetweenenvironmentalfactorsoperatingthroughepigeneticmechanismsandlate-onsetADhasreceivedrecentattentione.g.,Wuetal.,2008a.TheamyloidprecursorproteinAPPgeneislocatedonchromosome21locus21q21.3andcontains19exonsKangetal.,1987;StGeorge-Hyslopetal.,1987;Tanzietal.,1988.MultipleisoformsoftheproteinarecausedbyalternativesplicingYoshikai,1990.Inaddition,severalmutationsintheAPPgenehavebeenidentiedinfamiliesthroughpedigreeanalysis.Onevariantofearly-onsetADinvolvesavaline-to-phenylalaninesubstitutionatresidue717V717F;Murrelletal.,1991.Thismutation,knownastheLondon"variant,increasescleavageofAPPbygamma-secretase,resultinginhigherlevelsofA42.AnothervariantoffamilialAD,knownastheSwedish"variantMullanetal.,1992,involvesthedoublemutationK670Ni.e.,LystoAsnandM671LMettoLeu.Thisdoublepoint-mutationincreasescleavageofAPPbybeta-secretase,resultinginelevatedlevelsofbothA40andA42.Thepresenilin-1PS1genewaslocalizedtochromosome14locus14q24.3throughapedigreeanalysisof34individualswithearly-onsetADVanBroeck-29

PAGE 40

hoven,1992;Campion,1995;Alzheimer'sDiseaseCollaborativeGroup,1995,andspecicallydescribedbySherringtonetal.95.ThePS1geneconsistsof14exonsandthecodingregionisestimatedat60kbRogaev,1997;Del-Favero,1999.Presenilin-1isanintegralmembraneproteinthatcleavesNotch1Ikeuchi,2002,andmayplayaroleduringembryonicsomitogenesisKoizumi,2001.Severalpathogenicallelicvariantshavebeenfound,includingM146LMettoLeuSherringtonetal.,1995;Morellietal.,1998,M146VMettoValADCG,1995andM146IMettoIleJorgensenetal.,1996;Gustafsonetal.,1998.Inaddition,themutationE318GGlutoGlyresultsinapredispositiontoFADTaddei,2002.MutationsofPS1decreasetheageofADonsettotheearly40sand50sbyselectivelyincreasinggamma-secretasecleavageofpeptidesC99andC83,whichsubsequentlyincreasesbrainA42levelsSelkoe,2001.Thepresenilin-2PS2geneislocatedonchromosome1locus1q31-q42.ThePS2genecontains12exonsofwhich10arecodingregions.The448-residuepolypep-tideencodedbytheprimarytranscriptofPS2shares67%homologywithPS1Levy-Lahad,1996.BothPS1andPS2genessharestructuralandfunctionalsimilarity,andtheirassociatedproteinsareexpressedinthesameregionswithinmammalianneuronsKovacs,1996.LikePS1,severalpathogenicallelicvariantshavebeeniden-tied,includingN131IAsntoIle,M239VMettoValRogaev,1995,andD439AAsptoAlaLleo,2001.Tomita97hassuggestedthatthismutationresultsinincreasedamyloidplaqueformationthroughalteredAPPmetabolism.TheapolipoproteinEApoEgeneislocatedonchromosome19locus19q13.2.ApoEispolymorphic,andallelicvariantsofApoEconferdierentialpredispositiontohumanaging-associatedcognitivedisorders,includingmemoryimpairmentandAD.ThepresenceoftheApoE4variantisassociatedwithamorerapiddeclineinepisodicmemory,relativetoApoE2andApoE3Wilsonetal.,2002.TheApoE4variant30

PAGE 41

maybeassociatedwithincreasedriskforbothADCorderetal.,1993;Gurejeetal.,2006andCreutzfeldt-JakobdiseaseAmouyel,1994,inagenedose-dependentmanneri.e.,morecopiesofthegeneconfersgreaterrisk.InteractioneectsbetweenApoEandcellular-levelmicroenvironmentalvariableshavebeenobserved,aswell.StudiesinAlzheimer'stransgenicmice,forexample,suggestthatconcurrentpres-enceoflatentherpessimplexvirustype-1HSV-1inthebrainmayexacerbatetheriskassociatedwithApoE4MillerandFedero,2008,bypromotingmore-favorableconditionsforAD-associatedneuronaldegeneration.Mild-to-moderateADpatientsreceivingantidiabeticrosiglitazonetreatmentwhoareApoE4-negativeshowmodestcognitiveimprovement,whileApoE4-positivepatientsdonotRisneretal.,2006.Additionally,thedeleteriouseectsofApoE4maybemodulatedbytheexpressionofadditionalgenes.Usingagenome-widesinglenucleotidepolymorphismSNPanalysistoidentifylatentgeneticinteractions,Reimanetal.007foundanas-sociationbetweentheApoE4andGRB-associatedbindingprotein2GAB2geneswhichquadruplestheriskforlate-onsetADinApoE4-positiveindividuals.GAB2helpstoregulatetauphosphorylation,throughtheAKT/GSK3betakinasecascade.AlthoughtheroleofthetaugeneinAlzheimer'spathologyremainslargelyun-clear,evidencefromtransgenicmousestudiesdemonstratestheinteractionbetweentauandAPPexpressiononsubsequentpathologyandbehavior.Forinstance,al-thoughnochangeinbeta-amyloidlevelsorplaqueburdenwasdetectedintransgenicmicewithbothhumanAPPoverexpressionandreducedtauexpressioni.e.,eitherhomogeneousorheterogeneoustau-knockout,theseanimalsperformedcomparablytowild-typeanimalsintheMorriswatermazeRobersonetal.,2007.TheneuronalsortingreceptorSOR1geneisassociatedwithlate-onsetADRogaevaetal.,2007.SOR1promotesrecyclingofAPPfromthecellsurface,andunderexpressionofSOR1permitsbetaamyloidproductionfromAPPthrough31

PAGE 42

analternativepathway.Inaddition,apointmutationinamitochondrialenzyme-codingsequencehasbeenassociatedwithADLinetal.,1992,andgenesfromchromosomes10locus10p3Zubenko,2001and20Olsonetal.,2002havebeenimplicated.Epigenetics,thestudyofheritableregulationofgeneexpressionwithoutalter-ationoftheDNAsequenceWaddington,1953;Weinhold,2006;Dolinoyetal.,2007,isagrowingareainbioinformaticsresearchsupportedbygenome-widescansoffam-iliesBockandLengauer,2008.Epigeneticinheritancehasbeenproposedasonemechanismunderlyingnon-genotoxiccarcinogenesise.g.,peroxisomeproliferator-activatedreceptor;Gonzalez,2002,asthmaVercelli,2004,autismLopez-RangelandLewis,2006,andpossiblylate-onsetADBassettetal.,2002,2006.Addition-ally,modicationsofgeneexpressionduringearlymammaliandevelopmentcausedbynutritional,environmental,andevenbehavioralfactorscanprofoundlyaltertheresultingphenotypeDolinoyetal.,2007;epigeneticinheritanceexplains,inpart,howindividualswithidenticalgeneticmakeupmayexhibitdivergentphenotypictra-jectoriesoverthelifespanFragaetal.,2005.DierentpatternsofDNAmethylationbymaternalandpaternalallelesduringearlydevelopment,forexample,canresultingeneexpressiondeterminedbyparent-of-originimprinting"Lewin,2008,p.833.Twoextremelyrarehumangeneticdisorders,Prader-WillisyndromePWS;OMIM176270andAngelmansyndromeAS;OMIM105830,illustratethedisparatephe-notypiceectsofgenomicimprinting:Partialorcompletedeletionof15q11-q13i.e.,bands11through13ofthelongarmofchromosome15fromthematernalcopyofthechromosomeresultsinAS,characterizedbydelayedmotor,speechandcogni-tivedevelopment,seizures,andeccentricaectivebehaviorAngelmansyndromewasonceknown,pejoratively,ashappypuppetsyndrome"Magenisetal.,1987;similarcorruptionofthepaternalcopyresultsinPWS,characterizedbyhyperpha-32

PAGE 43

gia,hypotonia,andimpairedlearningabilityLedbetteretal.,1981.BecauseDNAmethylationeventscanoccurthroughouttheentirelifespanBjornssonetal.,2008,epigeneticmodulationofindividualphenotyperepresentsanongoingprocess.1.8AnimalModelsofAlzheimer'sDiseaseAnimalmodelsareusefulinresearchforidentifyingcharacteristicbiochemical,histopathological,andbehavioralmanifestationsofneurologicaldisorderse.g.,Gal-lagherandRapp,1997.Additionally,thesemodelscanbeusedtoelucidatethefunctionalroleofgenesorproteinsandtheimpactofgeneticmanipulationonpheno-typicexpression,aswellastoevaluatetherapeuticecacyofpharmacologicand/orbehavioralinterventions.Theassessmentandanalyticalmethodologiesforanimalmodelsshouldcloselyresemblethoseoftheirhumancounterparts,aswelle.g.,MRIinlivingtransgenicmice,Jacketal.2007.AnidealmodelforAlzheimer'sdisease,forexample,shouldmanifestprogressivecognitiveimpairmentandAD-characteristicneuropathologyJanusandWestaway,2001;Dodartetal.,2002;Bloometal.,2005;Brasnjevicetal.,2006.Inaddition,reproducibilityofresultsbydierentinvestiga-torsisnecessaryforestablishingavalidanimalmodel.Insomecases,naturally-occurringexamplesofAlzheimer's-likeneuropathologyalreadyexist,whichcanbestudiedasmodelsystemsofdiseaseNakayamaetal.,2004.Forinstance,cognitivedysfunctionsyndrome,ageriatricneurologicaldis-orderindomesticcatsLevineetal.,1987anddogsRuehletal.,1995;Cum-mingsetal.,1996a,1996b;Borrasetal.,1999,issymptomaticallyreminiscentofhumanAlzheimer'sdiseasee.g.,confusion,sensorydisorientation,reducedsocialinteraction,vocalization,disruptionofsleep-wakecycle,andurinaryincontinence.HistopathologicalchangessimilartohumanAD,includingextraneuronaldepositionofbetaamyloidprimarilyA42proteinandintraneuronalaccumulationofhyper-33

PAGE 44

phosphorylatedtauproteinalthoughnomatureneurobrillarytangles,havebeenobservedinaged0+year-oldcatsHeadetal.,2005;Cummingsetal.,1996c;Gunn-Mooreetal.,2006.ThehighestconcentrationsofCongophilicstainingwerefoundinthedeepcorticalandanteriorcerebralregions,andintenseneuronalAT8-immunoreactivitywasdetected.Inaged-18y/odogs,scattereddiusebutnotcompactbetaamyloidplaquesweredetectedinthecerebralcortexusingimmuno-histochemicalstaining,inadditiontoextensivecerebrovascularamyloiddepositionBorrasetal.,1999.Bothcatsanddogsrespondpositivelytoantioxidant-enricheddiets,withimprovedcognitiveperformanceCotmanetal.,2002;Milgrametal.,2005andincreasedlongevityCuppetal.,2006.Inadditiontodomesticcanineandfelinespecies,bothagednonhumanprimateschimpanzeesandorangutans;Gearingetal.,1994,1997andpolarbearsTekirianetal.,1996exhibitdiuseA42-specicplaquesinthecortexandhippocampus;tau-immunoreactivepairedhelicallamentshavealsobeenfoundinagedchimpanzeecorticaltissueRosenetal.,2008.BothneurobrillarytanglesandneuriticplaqueshavebeenfoundinthecerebralcortexofagedsheepNelsonetal.,1994.Inaddition,argyrophilicneurobrillarytangles,aswellasbothneuriticanddiuseplaques,werefoundinthecortexandhippocampusofanagedwolverineRoertgenetal.,1996.Transgenicanimals,bycontrast,representarticialconstructedmodelsystemsinwhichoneormoregenesaredeleted,mutated,and/oroverexpressedJaenisch,1988;Leeetal.,1996.Theresultinggenetically-engineeredanimalsareusefulsub-jectsforobservingandmeasuringthepathogenesisandprogressionofdisorders,aswellasforexaminingtheeectsofpreventiveortherapeuticinterventions.Forstudyingmanyhumanneurodegenerativediseases,includingAD,transgenicmicearecommonlyusede.g.,Emilienetal.,2000;Wongetal.,2002.ThesemiceareproducedthroughpronuclearinjectionofanappropriatecDNAtransgeneintoa34

PAGE 45

single-cellmouseembryowhichisimplantedintoapseudopregnantfemalemouse.Successfulincorporationofthetransgeneintothemousegenomeisveriedbygeno-typingtheospringbypolymerasechainreactionPCRatweaning.Usingthisprotocol,multiplelinesoftransgenicmousemodelshavebeendevelopedandmain-tainedforstudyingAlzheimer'sdisease.Theselectionofanappropriatebackgroundmousestrainisanimportantconsideration,howeverGerlai,1996;Carlsonetal.,1997;Crawleyetal.,1997;PicciottoandWickman,1998;Nguyenetal.,2000;Pughetal.,2004.InbredFVB/NorC57BL/6JlinescarryingtheAPPtransgene,forinstance,aremorepronetoprematuredeathprecededbybehavioralanomaliesneophobia,decreasedY-mazespontaneousalternationactivity,relativetooutbredhybridsCarlsonetal.,1997.InbredCBA/JmiceexhibitlearningimpairmentintheMorriswatermaze,whichmayberelatedtoaberranthippocampalelectrophys-iology,specically,impairedlong-termpotentiationintheCA1regionNguyenetal.,2000.1.8.1PDAPPTransgenicMouseModelThePDAPPtransgenicmousemodelincludesasingle-mutationhumanAPPgene,withphenylalaninereplacingvalineatresidue717V717F.ThehAPPmini-gene,usingahumanplateletderivedgrowthfactorbPDGF-bpromotor,isincor-poratedintomicehavingaC57B6xDBA/2F1hybridstrainbackgroundGamesetal.,1995;Rockensteinetal.,1995.PathologicalCharacterizationAsignicantincreaseintransgenicAPPlevelsisdetectedinthecerebralcortexandhippocampusbetween4and18monthsofageGamesetal.,1995,includinga17-foldincreaseinhippocampalbetaamyloidprimarilyAbeta[1-42]between4and8monthsofageJohnson-Woodetal.,1997.Atthreemonthsofage,signicant35

PAGE 46

hippocampalatrophyisdetectedDodartetal.,2000.At4-5months,enhancedpaired-pulsefacilitationPPFandrapidly-decayinglong-termpotentiationLTParenotedLarsonetal.,1999.Inaddition,decreasedproliferationofneuronsinthehippocampalsubgranularzoneSGZoccursduringtherstyearDonovanetal.,2006,suggestingthatalteredneurogenesiscontributestodecreasesinhippocampalfunctionintheseanimals.NeitherneurobrillarytanglesnorpairedhelicallamentshavebeendetectedinthePDAPPtransgenicmousemodelMasliahetal.,1996;Masliahetal.,2001.Althoughthree-tofour-month-oldhomozygousPDAPPmiceexhibitedsomeex-tentofmaturebetaamyloiddeposits,notallheterozygousPDAPPmicemanifestmaturedepositsDodartetal.,2000.Additionalstudieshaveshownthatbetweenfourtosixmonthsofage,nosignicantpathologyisobservedinheterozygotican-imalsGamesetal.,1995.However,betaamyloiddepositsarefoundexclusivelyinthehippocampus,corpuscallosum,andcerebralcortexoftransgenicanimalsinanage-dependentfashionbeginningatsixmonthsDodartetal.,2000.Eight-totwelve-month-oldheterozygousPDAPPmiceexhibitstructuralpathologicfeaturesreminiscentofhumanADMasliahetal.,1996,includingsimilaritiesinamyloiddeposition,dystrophicneurites,andglialcellreactivity.ThioavinS-positivebetaamyloiddepositsaredetectedinthehippocampusandentorhinalcortexbetween12and15months,withlittleornodetectabledepositsearlierReillyetal.,2003.By16-17monthsofage,increasedplaquedensitywasobservedwithintheentorhinalcortex,dentategyrus,andCA1ofthehippocampusChenetal.,2000,anddiuseplaqueswerefoundwithinthehippocampusandentorhinalcortex.AninteractionamonghumanAPP,ApoE,andACTexpressioninthePDAPPmousemodelhasbeenshowntoinuencebetaamyloidplaquedepositionandchar-acterNilssonetal.,2004.PDAPPmicelackingbothApoEandACThaveno36

PAGE 47

compactbetaamyloiddeposits,andonlyminimaldiusedeposits.OverexpressionofACTinApoE-knockoutPDAPPmiceincreasesdiusebetaamyloiddeposition,butcompactdepositsremainabsent.ExpressionofApoEproduceshigherdiusebetaamyloiddeposition,relativetoApoE-knockouts,andisassociatedwiththepresenceofcompactdeposits.ThepresenceofbothApoEandACTproducesthehighestlevelsofbothdiuseandcompactbetaamyloiddeposition.BehavioralCharacterizationBehavioralmanifestationsanddecitsprecedeamyloidplaquedeposits,withradialarmmazeimpairmentreferencememoryerrorsdetectedatthreemonths,comparedtowildtypecontrolsDodartetal.,1999.Objectrecognitionmemoryimpairmentisnotedatsixmonths,withsignicantdeclineidentiedat9-10monthsDodartetal.,1999;Dodartetal.,2000.Correlation-,factor-,anddiscriminant-analysesofbehavioralmeasuresobtainedfromcomprehensivetaskbatteriessupporttheexistenceofuniquelyidentiablesen-sorimotorandcognitivedomains,aswellasdomain-specicpathologicalcorrelationsinthePDAPPmousemodelLeightyetal.,2004.Signcantcorrelationsareob-servedbothwithinandbetweenwatermazetasksMorrismaze,platformrecogni-tion,andradialarmwatermaze,andbetweenthesetasksandCongophilicbetaamyloiddepositionwithinthecerebralcortexandhippocampusof15month-oldPDAPPmiceLeightyetal.,2004.Inparticular,spatialmemorytaskshaveiden-tieddomain-specicdecitsincognitiveperformance.Progressivedecitsinwatermazeserialspatialmemoryareseenbetween13and18months,andnon-progressivedecitsinwatermazespatialreferencememoryaredetectedbetween13and18months,andasearlyas3-4months,comparedwithnontransgenicmiceChenetal.,2000.37

PAGE 48

Thepatternofcognitiveimpairmentin15month-oldPDAPPmiceisinuencedbyhumanApoEandACTexpressionNilssonetal.,2004.NoimpairmentintheMorrismazeandradialarmwatermazetasksareobservedinPDAPPmicelackingbothApoEandACT.ExpressionofApoEresultsinsubstantialimpairmentinboththeMorrismazeandradialarmwatermazetasks.OverexpressionofACTinApoE-knockoutmiceresultsinmodestimpairmentintheradialarmwatermazetask,butnormalMorrismazeperformance.1.8.2APPswTransgenicMouseModelsAdoublemutationoftheAPPgene,occurringatresidues670K670Nand671M671L,knownastheSwedish"mutationisthebasisforseveraltransgenicmousemodels.Themodelsdierinstrainbackgrounds,whichproducesmarkeddier-encesincognitiveandsensorimotorphenotypiccharacteristics.TheFVB/Nmousestrain,forexample,exhibitssusceptibilitytohippocampalandcorticalapoptosisandastrogliosisMoecharsetal.,1996aswellasretinaldegenerationVinoresetal.,2003,andtheAPPswmodelhavingthisbackgroundisparticularlyvulnerabletohippocampalcelldeathMohajerietal.,2004.Sturchler-Pierratetal.97describethegenerationofanotherAPPswmodelcalledtheAPP23mouse,producedbyinsertionofthehAPP751cDNAintotheXhoIsiteofanexpressionvectorcontain-ingthemurineThy1.2glycoproteingeneandpromotor.TheTg2576mousemodelHsiaoetal.,1996;Chapmanetal.,1999introducesmutanthumanAPPthroughahamsterprionproteingene,PrP,promotoronaC57BL6xC57BL6/SJLstrainbackground.PathologicalCharacterizationYoungAPP23micemonthsexhibitsparsebetaamyloiddeposits,however,substantialplaquedepositionincludingdensedepositsareseenat24monthsin38

PAGE 49

diverseregions,includingtheneocortex,hippocampus,andthalamusSturchler-Pierratetal.,1997.BothheterozygousandhomozygousAPP23miceexhibitextensiveplaquedepositsbetween14-18monthsofageintheneocortexandhip-pocampusCalhounetal.,1998,althoughnosignicantneuronallosswaspresent.Additionally,noneurobrillarytangleswereidentiedthroughimmunostainingandBielschowskisilverstainingSchwabetal.,2004.Ninemonth-oldanimalswhichre-ceivedintrahippocampalinjectionsofhumanbetaamyloidextractatvemonthsofageexhibitextensivehippocampalplaquedepositionMeyer-Luehmannetal.,2006.Theinductionofbeta-amyloidogenesisatthisearliertimepointwithinasusceptiblehostdemonstratesthenucleantseeding"aspectofAlzheimer'spathologyRiek,2006.NeuropathologyintheTg2576mousewasrststudiedbyHsiaoetal.1996.IncreasedlevelsofA40werefoundasearlyastwomonthsofage,withave-foldincreaseby11to13monthsofage.ElevatedlevelsofA42werealsofound,witha14-foldincreaseby11to13monthsofage.Amyloiddepositswithinthefrontal,tem-poral,andentorhinalcortices,aswellasthehippocampusandcerebellumwerefoundin11to13month-oldmiceHsiaoetal.,1996.Similarly,brainlevelsofbothA40andA42increasedfrom4-8monthsofagethrough12months,withbetaamyloiddepositswithinthehippocampusandcortexbythelatertime-pointPraticoetal.,2001.By18monthsofage,Tg2576micehavewidespreadbetaamyloiddepositioninthehippocampusandneocortexPraticoetal.,2001.Benzingetal.999alsofoundextensivebrainbetaamyloiddepositsin18month-oldTg2576mice,predom-inantlywithinthetemporalandcingulatecortices,accompaniedbyastrocytesandmicroglia.Betaamyloiddepositswithinthecingulatecortex,entorhinalcortex,andhippocampalregionCA1weredetectedin16month-oldTg2576miceIrizarryetal.,1997.Positiveplaque-associatedastrocyteactivationwasfound,consistentwith39

PAGE 50

Benzingetal.99.KingandArendash02bnotedextensivesynaptophysinstainingofthehippocampusandneocortexin19month-oldTg2576mice,relativetoage-matchednontransgenicanimals;decreasedstainingwasobservedwithintheplaquecores,withincreasedstainingaroundtheperiphery.Additionally,Wester-manetal.002suggestedanassociationbetweenthepresenceofinsolublebetaamyloidandMorriswatermazeretentionperformancedecitsinTg2576miceover10monthsofage.Lesneetal.2006associatedextracellularaccumulationofasoluble56kDbeta-amyloiddodecamernamedbeta-amyloidstar56"withspatialmemorydecitsobservedin6to14month-oldTg2576mice.BehavioralCharacterizationAPP23miceexhibitspatiallearningandmemorydecitsintheMorriswatermazetaskKellyetal.,2003;Lalondeetal.,2002;VanDametal.,2003.Evidenceofcognitiveimpairmentprecedesplaquedeposition;heterozygousAPP23miceshowsignicantbetaamyloidplaquedepositionat6monthsofage,butdemonstrateimpairmentinboththeacquisitionandretentioncomponentsoftheMorriswatermazeasearlyas3monthsofageVanDametal.,2003.Lalondeetal.002notedimpairmentof16month-oldmiceintheacquisition{butnotretention{componentofMorriswatermazeperformance.Kellyetal.003obtainedsimilarndings,with3,18,and25month-oldAPP23miceshowingprogressivelyincreasinglatenciesinthesametask.Tg2576miceexhibitadierentpatternofbehavioralimpairmentfromAPP23mice.Holcombetal.1999foundnoimpairmentinsensorimotortasks,Y-mazeentriesoralternation,Morriswatermaze,orvisibleplatformperformanceof3month-oldTg2576mice.KingandArendash002aalsofoundnoimpairmentinY-mazeentries,Morriswatermaze,visibleplatform,andcircularplatformperformancein3month-oldmice.However,theynotedimpairmentinselectedsensorimotortasksin-40

PAGE 51

creasedopeneldactivityandpoorbalancebeamperformance,aswellasY-mazealternation.Hsiaoetal.96,bycontrast,didnotndY-mazealternationdecitsin3month-oldTg2576mice,butnotedtheseimpairmentsat10monthsofage,aswellasincreasedlatencyofMorriswatermazeacquisition.Tg2576miceexhibitim-pairednovelobjecthabituationandreducedreactivitytospatialnoveltyasearlyas7monthsofageMiddeietal.,2006,suggestiveofhippocampal-mediatedattentionaldecits.Circularplatformperformancedecitsweredetectedin7month-oldmice,relativetonon-transgeniclittermatesPompletal.,1999.Spatio-temporalcon-textlearningimpairmenthasbeenobservedin10-12month-oldTg2576miceGoodetal.,2007,reminiscentofhumanepisodicmemorydecitsi.e.,what,"when,"andwhere"aspectsofobjectpresentation.Holcombetal.99foundnoim-pairmentinMorriswatermazeorvisualplatformperformanceby9monthsofage,butnotedthatTg2576miceshoweddecreasedY-mazealternation,comparedwithnon-transgenicanimals.Bycontrast,KingandArendash02anotedasignicantincreaseinvisibleplatformlatencyat9months.ManifoldlearningandmemoryimpairmentMorrismazespatialacquisitionandretention,circularplatformescapelatency,RAWMworkingmemorywasalsofoundin9month-oldtransgenicmice,relativetoage-matchednontransgenicsArendashetal.,2006.Additionally,trans-genicmicedisplayedsignicantsensorimotorimpairmentbalancebeamandstringagilitytasksby14months,aswellasimpairedY-mazealternationby19months,relativetoage-matchednontransgenicanimalsKingandArendash,2002a.Hsiaoetal.996observedsignicantimpairmentinMorriswatermazeretentionbetween9and15monthsofage.Inaddition,impairedreversallearningodordiscriminationtaskwasobservedin6month-oldTg2576miceZhuoetal.,2007,withanimalsrequiringmoretrialstoreachcriterion,relativetoage-matchednon-transgeniccon-trols.41

PAGE 52

1.8.3APPsw+PS1TransgenicMouseModelTheAPP/PS1mousemodelincorporatestwohumanFAD-associatedmutationvariants:theTg2576APPswSwedishdouble-mutationandthepresenilin-1PS1M146LmutationHolcombetal.,1998.PathologicalCharacterizationThemoststrikingfeatureoftheAPPsw/PS1doubletransgenicmousemodelisitsacceleratedbetaamyloiddepositiontemporalprole,relativetotheAPPswmouse.Compactbetaamyloiddepositsarefoundasearlyas12-16weeksofageand,by24-32weeksofage,plaquessurroundedbyreactiveastrocytesareobservedHolcombetal.,1998.Takeuchietal.000foundbetaamyloidplaquedepositsintheneocortexand,toalesserextent,thehippocampus,ofthreemonth-oldtrans-genicmice.Bysixmonths,smalldiuseplaquesandlargercompactplaqueswereubiquitousinthecortex.NochangesweredetectedinhippocampalCA1neurondensitynorinsynaptophysin-associatedimmunoreactivity,relativetoage-matchedsingletransgenicandnon-transgenicmiceTakeuchietal.,2000.TheseresultsdierfromthoseofGordonetal.002,however,whofoundbetaamyloiddepositsinsixmonth-oldAPP/PS1mice,butnotinthreemonth-oldanimals.Thesedepositswerefoundmainlyinthehippocampusandfrontalandentorhinalcortices,inassociationwithbothreactiveastrocytesanddystrophicneurites.By15monthsofage,positivestainingforreactiveastrocytesincreasedthroughoutthebrain,particularlywithinthecerebralcortexandstriatumGordonetal.,2002.Jensenetal.005alsoreportedsignicantbeta-amyloidimmunostainingandCongoredstainingwithinthecortexandhippocampusin17-month-oldmice.Borcheltetal.997foundsubstan-tialbetaamyloiddepositioninthehippocampusandcerebralcortexofAPP/PS1mice,withage-relatedincreasesinplaqueburdenfrom9monthsto12monthsof42

PAGE 53

age.Asignicantnegativecorrelationbetweentotalbetaamyloidburdeninthehippocampusandfrontalcortex,andT1-T4acquisitiontrialerrorreductionintheradialarmwatermaze,underscorestheparalleltrajectoriesofneuropathologicalandcognitivedeteriorationGordonetal.,2001.Congophilicstaininginthefrontalcor-texissignicantlypositivelycorrelatedwithcognitiveimpairmentworkingmemoryerrorsintheradialarmwatermazetaskinAPP/PS1miceArendashetal.,2001;Gordonetal.,2001.Inaddition,nonage-dependentcorticalelectrophysiologicalanomaliesaredetectedinAPPsw/PS1miceWangetal.,2002,manifestedasre-ducedthetato6Hzandenhancedbeta4to27Hzandgamma8to40HzEEGactivity,underscoringthepathophysiologicaldistinctivenessofthisgenotype.BehavioralCharacterizationAPP/PS1miceupto9monthsofagedonotexhibitdecitsinsensorimo-tortasksHolcombetal.,1999.However,by6to9monthsofage,thesemiceshowincreasedactivityi.e.,numberofentriesandimpairmentinY-mazealterna-tion,comparedwithage-matchednontransgenicmiceHolcombetal.,1999.Addi-tionalstudiesbyArendashetal.2001utilizedamorecomprehensivebehavioraltaskbattery,andidentieddomain-specicpatternsofimpairment.At5-7monthsofage,APP/PS1micedidnotdierfromage-matchednontransgenicsineithersensorimotor-orcognitive-basedtasks,includingY-mazealternation,withtheex-ceptionsofY-mazeentriesAPP/PS1showedgreateractivityandbalancebeamAPP/PS1showedimpairment.Inalaterstudy,however,transgenicanimalsbe-tween4.5and6monthsofageexhibitedimpairmentinbothMorrismazeacquisitionandretention,aswellasRAWMworkingmemoryJensenetal.,2005,relativetoage-matchednontransgenicmice.Thisearlyevidenceofcognitiveimpairmentco-incideswithincipientplaqueformationwithinvulnerablebrainregionscortexandhippocampus.Tenmonth-oldanimalsexhibitsignicantlygreaterdurationwithin43

PAGE 54

{aswellasincreasedpercent-entriesinto{theopen-armsoftheelevatedplusmazePughetal.,2007.By15-17monthsofage,APP/PS1miceshowedincreasedac-tivitybothinopeneldandY-mazeentriesaswellasgeneralsensorimotorandcognitiveimpairment,comparedtoage-matchednontransgenicanimalsArendashetal.,2001.Ethelletal.06notedworkingmemoryimpairmentoverallnumberoferrorsduringTrials4and5intheradialarmwatermazetaskineightmonth-oldAPP/PS1mice,comparedwithage-matchednon-transgeniccontrolanimals.1.8.4APPsw+PS1+TauTransgenicMouseModelThetrigenic-ADmouseLaFerla,2006wasproducedbysubcloninganhAPP695cDNAfragmentwiththeSwedishdoublemutationK670N,M671L,andhumantauwiththeP301Lmutation,intothemurineThy1.2expressioncassette.Thesewereinsertedbyco-microinjectionintopronucleiofsingle-cellembryosfromPS1M146Vmutantknock-inmice.Currently,thismousemodelmost-closelymirrorsthepathologicalproleofhumanAlzheimer'sdisease,withsimultaneousexpressionofhumanmutantAPPandtauproteinsandtheappearanceoftheirrespectivehistopathologicalandbehavioralmanifestations.PathologicalCharacterizationTrigenic-ADxTgADmicedevelopbothamyloidplaquesandneurobrillarytanglesinanage-relatedprogressivefashionBillingsetal.,2005.Intraneuronalbetaamyloidproteinisfoundinthehippocampusandamygdalaoffourmonth-oldanimals,butneitherplaquenortangleformationisevidentatthisageBillingsetal.,2005.Betaamyloiddepositsinitiallyappearinthecortex,asearlyassixmonthsofage,andspreadtothehippocampus,whiletaupathologyemergesinthereverseorderOddoetal.,2003a.Betaamyloiddepositionprecedestanglepathology,despitecomparableoverexpressionoftherespectivehumanmutanttransgenesOddoetal.,44

PAGE 55

2003a,andconsistentwiththeamyloidcascadehypothesisofAlzheimer'sdiseasepathogenesis.Synapticdysfunctionprogressesinanage-relatedmanner,withdecitsinlong-termpotentiationprecedingplaqueandtanglepathology,anddecitsinlong-termsynapticplasticityrelatedtotheaccumulationofbetaamyloidwithinneuronsOddoetal.,2003b;LaFerlaandOddo,2005.Inaddition,thetrigenic-ADmousemodelprovidesevidencethatoligomerizationofbetaamyloidinitiallyoccursintraneuronallyOddoetal.,2006.BehavioralCharacterizationAt2.5monthsofage,3xTgADmiceareprepathologicandcognitivelyunimpairedMorriswatermaze,althoughtheyexhibitinitiallydecreasedverticalopeneldac-tivitynumberofrearings,relativetoage-matchednontransgenicsGimenez-Llortetal.,2007,andbysixmonthsofage,increasedhorizontalopeneldactivityloco-motionisobserved.Byfourmonthsofage,decitsinlong-termmemoryretentioncuedMorriswatermazearefoundBillingsetal.,2005.Immunotherapy-inducedclearanceofbetaamyloid,whichaccumulatesinthehippocampusandamygdalabyfourmonthsofage,temporarilyalleviatescognitiveimpairment,however.Ad-ditionally,sixmonth-oldtrigenic-ADmiceexhibitpooreracquisitionintheMorriswatermaze,aswellassubsequentmemoryretentiondecitsprobetrial,relativetonontransgenicsGimenez-Llortetal.,2007.1.8.5AnimalModels:ACodaThemostcommonlyusedvariantsoftheAlzheimer'stransgenicmousemodelweredescribedinthissection.However,theseexamplesrepresentonlyasamplefromthegrowingdiversityoftransgenicmousestrainsavailabletoresearchersforexploringvariousaspectsofAlzheimerpathology.Transgenicmicearecurrentlythebestanimalmodelsystemforinvestigatingthepathogenesisandprogressionof45

PAGE 56

Alzheimer'sdisease,aswellasforevaluatingexperimentaldiagnosticandtherapeuticinterventions.However,theAlzheimer'stransgenicmouseisnotaperfectmodel,andisincompletewithrespecttopathology.Theimmuno-neuroinammatoryresponse,forinstance,diersbetweenhumanADandthemousemodel,possiblyduetoreducedsensitivityofmurinecomple-mentfactorstohumanbeta-amyloidWebsteretal.,1997.Inhumans,signicantmicroglialactivationandincreasedlevelsofcomplementfactorsareassociatedwithplaquecoresSchwabetal.,2004.Bycontrast,aweakermicroglialresponsethatislargelyconnedtotheplaqueperipheryisobservedintheAlzheimer'stransgenicmouseSchwabetal.,2004.Inaddition,plaque-associatedneuronallossoccurringinhumanADisnotconsistentlyobservedintransgenicmice.InTg2576mice,forexample,somestudiese.g.,Tomidokoroetal.,2001reportneuronallossassoci-atedwithbeta-amyloidplaques,albeitlimited,"whileotherstudiese.g.,SteinandJohnson,2002donot.Finally,itisnoteworthythathumanAPPdiersfromthemurinehomologbyonlyseventeenaminoacidsJankowskyetal.,2007andthat,whilewild-typerodentsdonotexhibitage-associatedAlzheimer-likeamyloidlesionsShiversetal.,1988forunknownreasonsSelkoe,1989;Caietal.,2001;Jankowskyetal.,2004,theco-expressionofmurineAPPinAlzheimer'stransgenicmicealtersthesolubilityanddistributionpatternoftheaggregatesJankowskyetal.,2007.Despitethese,andother,potentialdecienciesandlimitations,eachnewgenerationofAlzheimer'stransgenicmousemodelconstitutesanevolutionarystepofscienticprogress,bothreectingourcurrentknowledgeofAlzheimer'sdiseaseandpromisingnewinsights.Currently,observingandmeasuringtheovertbehavioralcognitivemanifesta-tionsofAlzheimer'sdiseaseprovidethemostdirect,andleastinvasive,meansforinvestigatingthedisorder.Thedevelopmentoftechnologyforobserving,measuring,46

PAGE 57

recording,classifying,andinterpretinganimalbehaviormirrorsadvancesinpsychol-ogy,neurology,andassociatedbehavioralsciences.Themodernresearchercanselectfromawiderangeofexperimentalapparatus,includingarenas,mazes,andoperantchambers,andutilizecomputer-mediatedsensorsfordetectingeventhemostsubtlesubjectresponses.Notsurprisingly,thetoolsavailablefordataanalysisreectthesophisticationofthesedevices,frombasicstatisticstocomplexcomputationalal-gorithms.Onlyrecently,however,haveneuroscientistsbeguntoexploitthegreaterpotentialofcontemporaryanalyticalenginesforbehavioralresearch.1.9TreatmentsforAlzheimer'sDiseaseSeveralclassesofpharmacologicagentsSuhandChecler,2002;CummingsandZhong,2006;Klafkietal.,2006;Silvestrellietal.,2006havebeendevelopedforthetreatmentofAlzheimer'sdisease,includingcholinesteraseinhibitors,N-methyl-d-aspartateNMDAreceptorantagonists,andsecretase-modulators.Inaddition,variousnaturalproductsareunderinvestigationastherapeuticagentsCoxandBalick,1994.1.9.1PharmaceuticalsCholinesteraseinhibitorse.g.,tacrine,donepezil,galantamine,andrivastigminehelpcompensatefordiminishedavailabilityoftheneurotransmitteracetylcholinebyblockingtheactionofacetylcholinesterase,whichnormallybreaksdownacetylcholinefollowingreleaseGauthier,2001.Theresultingincreaseinsynapticconcentrationofacetylcholinemodestlyamelioratesthedepletionoftheneurotransmittercausedbyprogressivelossofcholinergicneurons,butdoesnotaltertheunderlyingpatho-geneticmechanismofthedementia.TheseagentshavedemonstratedeectivenessinmildtomoderateADHolmesetal.,2004,albeitforlimitedduration-2yrs.47

PAGE 58

TacrineCognexTMwastherstcentrally-actingacetylcholinesteraseinhibitorap-provedforthetreatmentofAlzheimer'sdisease.Clinicaltrialsinindividualswithmildtomoderatedementiashowedimprovementsinlanguageskillsproduction,comprehension,andwordrecognitionRaskindetal.,1997,anddemonstratedre-liefofmajorpsychiatricsymptomsanxiety,apathy,andhallucinationsinindivid-ualswithmoderatedementiaFarlowetal.,1992;Knappetal.,1994;Kauferetal.,1996.Tacrinehasbeenreplacedbynewermedications,duetoitspoororalbioavail-abilityandadversedrugreactionissuesincl.gastrointestinalandurinaryproblems,hepatotoxicityQizilbashetal.,1998.DonepezilAriceptTM,Eisai,whichwasin-troducedin1997,features100%oralbioavailability,fewerreportedsideeects,andalongerhalf-lifei.e.,requiresfewerdailydoses,comparedwithtacrineScarpinietal.,2003.ClinicaltrialsinmildtomoderateADpatientsshowedimprovedcognitionandactivitiesofdailylivingWinbladetal.,2001,andreducedpsychiatricsymp-tomsanxiety,delusionsinmoderatetosevereADpatientsRogersetal.,1998;Feldmanetal.,2001.GalantamineRazadyneTM,Ortho-McNeilNeurologicsisanalkaloidoriginallyisolatedfromowersandbulbsoftheVoronovsnowdropGalan-thusworonowii;AmaryllidaceaeScottandGoa,2000.Galantamineisnotonlyacompetitivereversibleacetylcholinesteraseinhibitor,butalsoincreasesacetylcholinereleasebymodulatingnicotiniccholinergicreceptorsWoodru-Paketal.,2001.InmildtomoderateADpatients,galantaminehasbeenshowntoimprovecognitionanddecreaseanxietyandhallucinationsTariotetal.,2000.RivastigmineExelonTM,Novartisinhibitsbothacetylcholinesteraseandbutyrylcholinesterase,andispartic-ularlybenecialinpatientsexhibitingAlzheimer's-orParkinson-associateddelusionsorhallucinationsBurnetal.,2006;Gauthieretal.,2006;Touchonetal.,2006.Ri-vastigminebecametherstproductapprovedforthetreatmentofmildtomoderatedementiaassociatedwithParkinson'sdiseaseEmreetal.,2004in2006.48

PAGE 59

NMDAreceptorantagonistse.g.,memantinebindtoNMDAreceptorswithgreateranitythanmagnesiumion,thuspreventingtheprolongedinuxofcalciumionsunderlyingneuronalglutamatergicexcitotoxicityobservedinADCacabelosetal.,1999.Theseagentscandissociatefromthepostsynapticreceptorstopermitglu-tamatebinding,thuspreservingphysiologicalandsignal-conductiveprocessesofthereceptor.MemantineNamendaTM,ForestPharmaceuticals,introducedin2003,isanon-competitive,low-anityNMDAreceptorantagonistLipton,2005,andtherstpharmaceuticaltherapeuticagentforADwhichtargetstheglutamatergicsys-temRobinsonandKeating,2006.ItisindicatedfortreatingmoderatetosevereADReisbergetal.,2003and,whenadministeredwithdonezepil,shownsenhancedeectivenessforimprovingcognitivefunctioninsevereADTariotetal.,2004.Eightmonth-oldAlzheimer'stransgenicmiceAPP/PS1receivingdailymeman-tine0mg/kg/dayp.o.for2to3weeksshowedless-impairmentinMorriswatermazeacquisitioni.e.,reducedescapelatency,relativetountreatedtransgenicani-malsMinkevicieneetal.,2004,suggestingimprovedhippocampal-mediatedspatiallearningability.MemantineisalsoapotentserotonergicHT3receptorantagonistRammesetal.,2001,buttheclinicalrelevancetoADremainsunknown.SelectiveamyloidloweringagentsSALA,suchasR-urbiprofenFlurizanTM,tarenurbil,MyriadGenetics,representanewclassofpharmacologictherapiesfortreatingmildADcurrentlyunderinvestigation.TheseagentsarerelatedtoNSAIDs,andhavebeenshowntoreducethelevelsofpathogenicA42bothinvitroandinanimalmodelsMoriharaetal.,2002;Eriksenetal.,2003bymodulatingtheactionofgamma-secretase.BothR-ibuprofenandR-urbiprofenarepoorcyclooxygenaseinhibitors,buteectivelyreduceA42productioninvitroMoriharaetal.,2002.However,an18-monthPhase3clinicalstudyofFlurizanTMinpatientswithmildAlzheimer'sdiseaseshowednosignicantdierencebetweentreatedanduntreated49

PAGE 60

patients,withrespecttocognitive-ordailyliving-basedindiceseithercognition-anddailyliving-indicesGreenetal.,2008.1.9.2NaturalProductsSeveralnaturalproducts,includingmedicinalplantextracts,areunderinvestiga-tionasnootropicsandpotentialtherapeuticagentsforADbasedonethnobotanicaltraditionsPerryetal.,1999;Philipson,2003.Forexample,avonoidandter-penetrilactoneextractsfromtheMaidenhairtreeGinkgobilobaL.;Ginkgoaceaemayslowtheprogressionofdementia,andarebeingstudiedaspotentialtreatmentsforcognitiveimpairmentandAlzheimer'sdiseaseZimmermannetal.,2002;Cohen-Salmonetal.,1997;Kanowskietal.,1996;LeBarsetal.,1997;DeKoskyetal.,2006.TheantioxidantpropertiesofGinkgoextractsmayreduceAlzheimer's-relatedox-idativestressChristen,2000;Philipson,2003.DailyoraladministrationofGinkgoextractmg/kg/dayineightmonth-oldAlzheimer'stransgenicmiceTg2576forsixmonthsresultedinMorriswatermazeprobetrialmemoryretentionperfor-mancecomparabletoage-matchednon-transgenicanimalsStackmanetal.,2003.TherewasnosignicantspatiallearningdierencebetweenGinkgo-treatedandun-treatednon-transgenicanimals,howeverStackmanetal.,2003.StudiesinratsShifetal.,2006suggestthat,whileoraladministrationofGinkgoextractmaynotaecteitherreferencememoryorworkingmemoryintheMorrismazeandeight-armradialmazetasks,chronicadministrationimprovestheoverallrateofspatiallearn-ingovertime.OraladministrationofCelastrusseedCelastruspaniculatusWilld.;CelastraceaeoilimprovesmemoryretentionperformanceinratsNalinietal.,1995usingatwo-compartmentpassiveavoidanceparadigmBuresandBuresova,1963.Treatedanimalshavesignicantlylowerbrainlevelsofmonoamineneurotransmit-tersi.e.,norepinephrine,dopamine,andserotoninandtheirmetabolites,suggesting50

PAGE 61

boththatCelastrusoildecreasestheturnoverofcentralmonoaminesandthattheseneurotransmittersexertaninhibitoryinuenceonlearningandmemoryNalinietal.,1995.Inaddition,water-solubleextractsfromCelastrusseedshavebeenshowntoinhibitNMDAreceptorsinvitro,whichmayserveaneuroprotectiverolebypre-ventingglutamate-inducedneurotoxicityGodkaretal.,2004.HuperzineA,analkaloidisolatedfromtheChinesemedicinalherbQianCengTaHuperziaserrata[Thunb.]Trev.=Lycopodiumserratumandrelatedgeneraofclubmossferns,in-hibitsacetylcholinesteraseactivityandincreasesbrainacetylcholinelevelsforseveralhoursChengandTang,1998;Xiaoetal.,2000;Baietal.,2000;Zangara,2003.Vinpocetine,isolatedfromMadagascarperiwinkleVincaalba,increasescerebralbloodowinpatientswithsenilecerebrovasculardiseaseBalestrerietal.,1987.ThespiceturmericCurcumalonga,foundincurrypowder,containscurcumin,whichhasbeenshowntoinhibittheformationofbetaamyloidbrilsandtodestabi-lizepreformedbrils,invitroOnoetal.,2004.DailycurcuminreducesbothsolublebetaamyloidandplaqueburdeninTg2576APPswmice,withoutalteringAPPlevelsLimetal.,2001.Thereducedage-adjustedprevalenceofAlzheimer'sdiseaseobservedinIndiaGangulietal.,2000mayberelatedtodietaryturmeric.Youngmonth-oldAPP/PS1micereceivingextractofBacopamonnieratraditionalanti-aging,memory-enhancingtherapyfromIndiafor8monthsshowdecreasedcor-ticalA40andA42levels,relativetountreatedcontrolanimalsHolcombetal.,2006.TransgenicmiceTg2576receivingorally-administeredpomegranatejuiceshowedenhancedwatermazeperformanceanddecreasedbeta-amyloidbothsolu-bleandcompactdepositforms,relativetountreatedcontrolsHartmanetal.,2006.Cannabinoidderivativeshavebeenshowntoblockbetaamyloid-inducedmicroglialactivationinvitro,andmayhaveadditionalneuroprotectivepotentialRamirezetal.,2005.51

PAGE 62

1.9.3Behavior-BasedTherapiesAsanadjuncttopharmacologictreatments,severalbehavioralandpsychosocialinterventionandrehabilitationstrategieshavebeensuggestedforearlytomoder-ateAlzheimer'sdiseaseCohen-Manseld,2001;Olazaranetal.,2004;Sitzeretal.,2006.ResearchwithAlzheimer'stransgenicmice,whichexhibitprogressiveneuropathologyandcognitiveimpairmentreminiscentofhuman-AD,suggeststhatphysicalexercisemayalleviateAD-associatedcognitiveimpairmentNicholetal.,2007.Afterthreeweeksofadlibaccesstoanexercisewheel,aged16-18month-oldtransgenicanimalsperformedcomparablytoage-matchednon-transgeniccon-trolsintheradialarmwatermazetaskforbothshort-termworkingandlong-termreferencememoryperformanceNicholetal.,2007;bycontrast,sedentaryagedtransgenic-andnon-transgenicanimalswerereadilydistinguishable.StudiesinAlzheimer'stransgenicanimalmodelsdemonstratethecognitive-protectiveben-etsofenriched"i.e.,physically-andcognitively-stimulatinghousingconditions.Arendashetal.2004a,forexample,notedthataged16month-oldAPPswmicewhichareplacedintoenrichedhousingconditionsforfourmonthsexhibitcognitiveperformancespatialreferencelearning/memory,objectidenticationsignicantlysuperiortothatoflitter-mateshousedunderstandardlaboratoryconditions.Nodierenceswerefoundbetweenenriched-andstandard-housedmiceintotalAloadofeitherhippocampusorparietalcortexArendashetal.,2004a,suggestinganalternativemechanismtobrain-amyloidreductionfortheobservedcognitiveef-fects.Subsequentinvestigationshavefoundbothdecreasedcorticalbeta-amyloiddepositionLazarovetal.,2005;Ambreeetal.,2006andenhancedcognitiveperfor-manceCostaetal.,2007eectsofenvironmentally-enrichedhousingenvironments,relativetostandardanimalhousingconditions.Indeed,long-termcomplete"en-52

PAGE 63

richmentenvironmentscombiningsocial,physical,andcognitivestimulationmayprovideoptimalconditionstoprotectagainstcognitiveandfunctionalimpairment,andameliorateAD-associatedneuropathologyCracchioloetal.,2007.Further-more,behavioralinterventionsinvolvingphysicaland/orcognitivestimulationhavedemonstratedbenetsformaintenanceandevenmodestimprovementsinmentalfunctionofADpatients.AmbulatoryADpatientswhoparticipatedinacollectiveexerciseprogramonehour,twiceweeklyforoneyear;walking,strength-,balance-,andexibility-trainingregimeshowedaslowerdeclineinactivitiesofdailylivingADL;KatzIndexperformancemeasures,relativetononparticipantsRollandetal.,2007.QuayhagenandQuayhagen989showedthatADpatientswhoengagedinmentally-stimulatingactivitiesverbalandmemoryexercises,problem-solvingmaintainedcognitivefunctionduringthetreatmentperiod,comparedtopatientswhodidnotengageintheactivities.Indeed,ADpatientswhoreceiveregulartwo45-minsessions/day,threetimesperweekforvewkstrainingindailylivingactiv-itiese.g.,routinekitchenactivities,letter-writing,telephoneuseshowedsignicantimprovementsinbothcognitive-anddailylivingfunctionalityFarinaetal.,2002;bythreemonthsaftertrainingceased,however,thesebenetsvanishedandpatientsregressedtopre-trainingstatus.Insomecases,whereADpatientsdonotrespondtophysicaltherapeuticinterventions,thecausemayinvolveexecutivedysfunctionorgaitdisturbancesarisingfromAD-associatedcerebrovasculardiseaseScherderetal.,2007.53

PAGE 64

1.10StatementofPurposeTheinvestigationsdescribedinthisdissertationembracethemultidisciplinarycharacterofmodernAlzheimer'sresearchbyintroducing,implementing,andeval-uatinganovelhuman-mouseparallelcognitive/behavioraltestingparadigm,inad-ditiontocomplementaryanalyticalprotocolsbasedondataminingtechniques,forneurobehavioralassessmentinAlzheimer'stransgenicmice.AninterferencetestingparadigmrecentlydevelopedforevaluatingADpatientswasadaptedformouse-basedtestingusingspatialmemory-domainelementsinplaceoftheoriginalhumanse-manticlearningcomponents.Theeectivenessofthisnovelcognitiveassessmenttaskfordistinguishingtransgenicand/ortreatmenteectswasexaminedinseveralstudies.Dataminingmethods{widelyusedinengineering,businessandindustrialapplications{areonlyrecentlygainingrecognitioninneuroscienceascomplementaryanalyticmethodologiestowell-establishedstatisticalapproaches.Thesetechniqueswereshowntobeeectivetoolsforevaluatinggenotypeand/ortreatmenteectsusingmultimetricbehavioraldata.Thespecicaimsoftheseinvestigationswereto:1.Evaluatecognitiveeectsoflong-termcaeineadministrationinnontransgenicmice,usingbothconventionalstatisticalandcomplementarydataminingan-alyticmethods,inacomprehensivesensorimotorandcognitivetaskbattery.2.Examineshort-termcaeineadmininstrationinbothagednontransgenicandAlzheimer'stransgenicmice,toidentifypotentiallydierentialsensitivitytocognitive-protectivebenetsofcaeine.54

PAGE 65

3.Analyzethehumansemanticinterferencetaskdatasetusingconventionalsta-tisticalandcomplementarydataminingtechniques,toevaluatethediagnosticutilityanddierentialsensitivityoftheseresearchtoolsforhumanclinicalapplications.4.Demonstrateutilityofthenovelhuman-mouseparalleltestingparadigmin-terferencetask"fordistinguishingamongnontransgenicandAlzheimer'strans-genicmicewithandwithoutG-proteincoupledreceptorkinase-5GRK5gene-knockoutmanipulation.5.Utilizethenovelinterferencetasktoevaluatetherapeuticecacyofgranulo-cytemacrophagecolony-stimulatingfactorGM-CSFinbothnontransgenicandAlzheimer'stransgenicmice.55

PAGE 66

CHAPTER2BEHAVIORALASSESSMENTINALZHEIMER'STRANSGENICMICEComprehensivetestingbatterieshavebeendevelopedforphysicalandbehavioralassessmentofanimalsubjectsinlaboratorysettings.Thetheoreticalandpracticalnecessityofutilizingmultiplebehavioraltasksand/ormeasurescannotbeoverstatede.g.,Rogersetal.,1999;vanderStaayandSteckler,2001;Arendashetal.,2004b;Caeyenberghsetal.,2006;Vekovischevaetal.,2007,withjusticationsinclud-ingconvergentvalidation,enhanceddiscriminativecapacity,andbroadersamplingacrossthebehavioralrepertoire.Indeed,asCrabbeandMorris2004argue,usingtheexampleofarodentbehavioralmodelofhumanintoxication,severalmeasuresfromdierenttasksarenecessarytodepictstrain-independentsensorimotorman-ifestationsofethanolingestion.Inaddition,complexbehavioralphenomenae.g.,memoryandlearninghavebothspatialandtemporaldimensionswhichmustbead-dressedduringbothmeasurementandanalysis.Forexample,cognitiveimpairmentismoreeasilyidentied,relativetonormalcognitiveability,usingtaskswhichmanip-ulatethelatencytimedelaybetweenlearningandsubsequentrecallofinformationMcDonaldandOvermier,1998.AnearlyassessmentbatteryIrwin,1968consistedofftyobservationalcat-egoriese.g.,locomotoractivity,gripstrength,rightingreex,individuallyscoredusinganine-point-8scale.Thisscreenwasusedinpharmaceuticalresearchforevaluatingandcharacterizingdrugreponsesinanimals.Anothertestingbat-56

PAGE 67

teryMoseretal.,1995,developedbybehavioralneurotoxicologistsattheU.S.EnvironmentalProtectionAgency,classiesbehavioralobservationsintosixneu-rologicaldomains:Activity,autonomic,excitability,neuromuscular,physiological,andsensorimotorfunction.SmithKlineBeechamPharmaceuticalsdevelopedacom-prehensivetestingprotocolSHIRPA;Rogersetal.,1997encompassingbehavioralobservations,sensorimotorandneurophysiologicalprocesses,andcognitivefunction-ing.SHIRPAisusefulforidentifyingstrain-specicbehavioralpatternsbehavioralphenotyping"andfordistinguishingamonginbredmousestrainsRogersetal.,1999;Crawley,1999.Theprotocolsandmeasuresdescribedbelow,andsummarizedinTable2.1,com-prisethebehavioralassessmentbatterydevelopedbyGaryArendashandcolleaguesattheUniversityofSouthFloridaforexaminingtreatment-andtransgenicity-eectsinAlzheimer'stransgenicmice.Thetasksarepresentedinorderoftestingsequenceandincludepracticaldetails,aswellashistoricaldesignperspectives.2.1SensorimotorTasksandAssociatedMeasures2.1.1OpenFieldActivityTheapparatusconsistsofanopenbox,80cmby80cmsquarewith28cm-highwalls,paintedblack.Theooroftheboxisdividedintosixteenregionsbyagridofwhiteequispacedlines.Eachanimalisplacedinthecenteroftheoor,andthenumberoflinescrossedduringasingleve-minutetrialisrecordedOF.TheearlyversionofthischamberBroadhurst,1961featuredasomewhatlarger0cmx120cmx45cm,25-celloorgridenclosureinwhichrodentbehaviorcouldbeobservedforaspeciedtimee.g.,5min,1hr,andthenumberoflinescrossedorregions57

PAGE 68

enteredisrecorded.Theopeneldactivitytaskprovidesameasureofspontaneouslocomotorandexploratorybehavior.Circulararenase.g.,Hall,1934arealsoused.Inaddition,automatedmonitor-ingphotocell-detector,videotaperecordingmethodsaresometimesusedinplaceofahumanobservere.g.,Kafka,2003.Theoriginalparadigmofobservingnaiverodentand,subsequently,otherspeciesbehaviorinconnedarenasfocusedonau-tonomicresponsese.g.,defecation,urinationasindicesofemotionality,"anxietyandstressHall,1934;Hall,1936.Indeed,familiarityhabituationwiththeopeneldthroughrepeatedtestingdecreasesanxiety-likeresponsesinratsOssenkoppetal.,1994.Furthermore,thecognitive-emotionaldeterminantsofovertbehaviorhavebothsituationalenvironmentalanddispositionale.g.,genetic,physiologicalcomponentswhichremainlargelyunexploredRamosandMormede,1998.2.1.2BalanceBeamAwoodenrodmeasuring50cmlongby1cmwideisxedbetweentwocolumnarsupports,suspended45cmaboveapaddedtablesurface.Ateachendoftherod,mountedatopthesupportingcolumn,isa14cmby10cmescapeplatform.Atthestartofeachofthree60-secondtrials,themouseisplacedatthecenterofthenarrowrod,alignedperpendicularlytotherodslength.Theamountoftimetheanimalremainsontherodbeforefalling,upto60sec,isrecorded.Iftheanimalreacheseitherplatformescapes,themaximumtimeof60secisassigned.Theaveragetimeacrossthreetrialsisrecorded,aswellBB.Aseriesofprogressively-narrowerbeamse.g.,Carteretal.,1999canbeusedforquantifyingsensorimotordecits.Thebalancebeamtaskreectsoverallmotorcoordinationandbalance.58

PAGE 69

Table2.1.Task-associatedbehavioralmeasuresusedinanalyses Task Symbol Description OpenField OF Numberoflinecrossings BalanceBeam BB Averagelatency,inseconds StringAgility SA Performanceratingscore Y-Maze YM-AE NumberofY-mazearmentries YM-PA Percentageofspontaneousalternations ElevatedPlus EP-CE Numberofclosed-armentries Maze EP-OE Numberofopen-armentries EP-TO Timespentinopenarms,inseconds WM-Avg Overallaverageescapelatencyacrossdays duringtheacquisitionphase,inseconds MorrisWater WM-Fin Averageescapelatencyonthenaldayofthe Maze acquisitionphase,inseconds WM-Ret Percentageoftimespentinformerplatformcontainingquadrantduringretentiontesting CPE-Avg Overallaveragenumberoferrorsacrossdays Circular CPE-Fin Averagenumberoferrorsonthenalday Platform CPL-Avg Overallaverageescapelatencyacrossdays,insec. CPL-Fin Averageescapelatencyonthenalday,insec. Platform PR-Avg Overallaveragelatencyacrossdays,insec. Recognition PR-Fin Averagelatencyonthenalday,insec. RME-T4 AveragenumberoferrorsduringTrial#4, acrossallblocks RME-T5 AveragenumberoferrorsduringTrial#5, acrossallblocks RME-FT1 AveragenumberoferrorsduringTrial#1 ofnalblock RadialArm RME-FT4 AveragenumberoferrorsduringTrial#4 WaterMaze ofnalblock RAWM RME-FT5 AveragenumberoferrorsduringTrial#5 ofnalblock RML-T4 AveragelatencyforTrial#4acrossallblocks,insec. RML-T5 AveragelatencyforTrial#5acrossallblocks,insec. RML-FT1 AveragelatencyforTrial#1ofnalblock,insec. RML-FT4 AveragelatencyforTrial#4ofnalblock,insec. RML-FT5 AveragelatencyforTrial#5ofnalblock,insec. 59

PAGE 70

2.1.3StringAgilityTheapparatusforthistaskresemblesthatoftheBalanceBeamtask,withthewoodenrodreplacedbyataut,cottonstring.Atthestartofasingle60sectrial,themouseispositionedmidwayalongthelengthofthestringandallowedtograspthestringwithitsforepaws,andthenreleased.Theanimal'sperformanceisassessedusingasix-pointscoringsystem,andthismeasureisrecordedSA:0"isassignedifthemousefallsfromthestring,"isassignedifthemousemaintainsitsgriponthestringfor60secusingitsforepaws,2"isassignedifthemousemaintainsitsgriponthestringfor60secusingitsforepawsandeitherhindlimb,"isassignedifthemousemaintainsitsgriponthestringfor60secusingallfourlimbs,4"isassignedifthemousemaintainsitsgriponthestringfor60secusingallfourlimbsanditstail,and,"isassignedifthemouseescapesbyreachingeitherplatform.Asecondtrialispermittediftheanimalfallsfromthestringimmediatelyafteritisreleasedatthestartofthersttrial.Thistaskextendsanearlytechniqueformeasuringgripstrengthwhichutilizedthewiregridcoveroftheanimal'scage:themousewasallowedtograspthewiregrid,whichwasimmediatelyinverted,suspendingtheanimalseveralinchesabovethecageoore.g.,Sangoetal.,1996.Thestringagilitytaskprovidesameasureofforepawgripcapacityandoverallphysicalstrengthandagility.2.2CognitiveTasksandAssociatedMeasures2.2.1Y-MazeTheapparatusconsistsofaY-shaped,three-armedcmlongby4cmwidemazeenclosedby40cm-highwalls.Allvisiblesurfacesoftheinteriorarepaintedblacktoprovideauniform,neutraltestingenvironmentwithoutspatialorientation60

PAGE 71

cues.Alightsourceispositionedabovethecenterofthemazewherethethreearmsintersect.Themouseisplacedinthecenterofthemazeandthesequenceofarm-visitsisobservedforasingle,ve-minutetrial.ThetotalnumberofarmentriesYM-AEisrecorded,asameasureofexploratorybehavior,aswellasthepercentagespontaneousalternationYM-PA;i.e.,theratioofthenumberofvisitsofallthreearmsduringthreeconsecutivearm-choicestothetotalnumberofarm-entries,whichisassociatedwithgeneralmnemonicfunctionLamouretal.,1989.TheY-MazeparadigmisavariationoftheT-Maze,anearlierdesigninwhichworkingmemoryandexploratorybehaviorareexaminedinmoderatelyfood-orwater-deprivedtestsubjectsintroducedintothelowerendofaT-shapedmazeinwhichasmallamountoffoodorwaterhasbeenplacedatbothend-wallsofthelongercross-arm.Duringthepre-testinghabituationperiod,theanimalmustlearntoalternatearm-visitstoobtainthereinforcer,oftenrequiringseveralweeksoftrainingHepleretal.,1985;Markowskaetal.,1989.ThedetrimentaleectsofdisruptingcholinergiccircuitsofthebraineitherbynucleusbasalislesionsoracetylcholinereceptorantagonistsonworkingmemoryhavebeendemonstratedusingT-mazedelayedalternationmeasuresHepleretal.,1985;Mastropaoloetal.,1988.TheY-mazetaskinvolvesbothcognitivebasicmnemonicfunctionandsensorimotorexplorationcomponentsofbehavior.2.2.2ElevatedPlusMazeTheElevatedPlusMazeisafour-armedmazeconsistingoftwooppositeopen"armsandtwooppositeclosed"arms,eachmeasuring30cmlongby5cmwide.Thefourarmsmeetatrightangles,forminga5cmby5cmcenterregion.Additionally,theclosed"armsaresurroundedby15cm-highwalls.Alightsourceispositionedabovethecenterofthemazewherethefourarmsintersect.Theentireapparatus61

PAGE 72

ispositioned80cmabovetheground,andtheinteriorwallsarepaintedblacktoreduceenvironmentalcues.Themouseisplacedinthecenterofthemaze,facingaclosedarm,andobservedforasingle,5-minutetrial.ThenumberofopenarmentriesEP-OEandclosedarmentriesEP-CE,aswellasthetotaltimespentintheopenarmsEP-TO,isrecorded.ThisparadigmisbasedontheapparatusandinterpretationdescribedinBrileyetal.986,downscaledfromtheoriginaldimensions0cmlongby10cmwidearmssuitableforrats.Brileyetal.86alsorecordedthetotalnumberofarmentriesduringasingleve-minutetrialasameasureoftotallocomotionandthepercentageofopen-armentriesasanindexoffearresponse,andfoundthatratsshowamarkedpreferenceforenclosedarms.Briley'sdesign,inturn,wasbasedontheearlierob-servationthatratsexhibitreducedexploratorybehaviorandenhancedavoidanceofopenelevatedalleys,comparedwithcoveredtunnelsMontgomery,1955.Theeectsofanxiolyticchlordiazepoxide,pentobarbital,ethanolandanxiogenicFG-7142,caeine,picrotoxinagentshavebeenstudiedusingtheelevatedplusmaze,inconjunctionwithaholeboardtestLister,1987.Thistaskisusedtoevaluateemotionalityandgeneralanxiety,althoughabilitiesrelatedtodecision-making"andrisk-assessment"mayberepresentedaswellRodgersandJohnson,1995.Conr-matoryfactoranalysese.g.,WallandMessier,2000,however,suggestatwo-factorsolutionforthistask.Indeed,theLong-TermCaeineAdministrationinNon-transgenicMice"studyreportedinthisdissertationillustratessegregationbetweenthearm-residencydurationmeasureEP-TOandthetwoarm-entryfrequencymea-suresEP-OE,EP-CE,underscoringthedistinctionbetweenanxiety-associatedandexploratory-locomotormetricsintheelevatedplusmazetask.62

PAGE 73

2.2.3MorrisWaterMazeSubmergedPlatformTheapparatusconsistsofawater-lled,100cmdiametercircularpoolwhichisdividedintofourequal-sizedquadrantsbytwoperpendicularlinesdrawnontheoorofthepool.Aclear9cmdiametercircularplatformisplacedinthecenterofquadrant#2,submerged1.5cmbeneaththesurfaceofthewater.Assortedvisualcuesareplacedalongtheexternalcircumferenceofthepool,toserveasextramazenavigationallandmarkse.g.,Suzukietal.,1980.Thetrainingandtestingprocedureconsistsoftwophases:aten-dayacquisition"phase,involvingfourtrialsperday,followedbyaone-day,single-trialretention"phase,whichisvideotape-recordedforlateranalysis.ThetwophasesoftheMorrisWaterMazetaskinvolvespatiallearningacquisition"andreferencememoryretention"processes.Oneachdayoftheacquisitionphase,themouseisplacedintoeachofthefourquadrantsrandomizedorder,initiallyfacingthesidewall,andtheaveragelatencytolocateandmountthesubmergedplatformisrecordedmaximumof60secpertrial.Theanimalisallowedtoremainontheplatformhavingreachedtheplatformeitherbyswimmingorthroughgentleguidancebytheexperimenterfora30secintertrialstay"period.Themaximumlatency0secisrecordedforanytrialinwhichtheanimalrequiresguidancetoreachtheplatform.Theaveragelatencyforeachdayiscalculated,andboththenal-dayaveragelatencyWM-FinandoverallaveragelatencyacrossalldaysWM-Avgarerecorded.Onthedayfollowingcompletionoftheacquisitionphase,thesubmergedplatformisremovedandthemouseisplacedintothequadrantoppositetheformerlyplatform-containingquadrantforasingle,60secprobe"trial.Thevideotapeoftheprobetrialissubsequentlyexaminedtodetermineperformanceparameters:percentageoftotalswimtimespentinquadrant63

PAGE 74

#2WM-Ret,quadrantpreference,numberofannuluscrossings,swimpathand/oraveragespeed,etc.TheMorriswatermazeMorris,1981;Morris,1984,originallyintendedtoexam-ineneuroanatomicalsubstratesofspatiallearningandmemoryinrats,isthemostcommonparadigmforcognitiveevaluationinrodentsD'HoogeandDeDeyn,2001.HippocampallesionsinbothratsMorrisetal.,1982;Eichenbaumetal.,1990andmiceLogueetal.,1997,forexample,consistentlyproduceacquisitiondecitsinthistask.Bothspecies-andstrain-specicdierencesinperformancehavebeenreported,withratsgenerallysuperiortomiceintheMorriswatermazeWhishawandTomie,1996andC57BL/6strainperformingbetterthanSwissWebstermiceWrightetal.,2004.Idiosyncraticoranomalousbehaviors,suchascirclingorpreferenceforsinkingoverswimming,arecommonlyreportede.g.,Wahlstenetal.,2005.Inaddition,videorecordingequipmentisoftenutilizedtoincreasetestaccuracyandeciency.Forexample,recordedsessionscanbeviewedandevaluatedindependentlybymultipleobservers,thusreducingexperimenterbiasandimprovingthereliabilityofbehavioralresponsescoringGrazianoetal.,2003;TecottandNestler,2004.2.2.4CircularPlatformTheapparatusconsistsofawalled70cmdiameterarenawithsixteencircularescape"holesequidistantly-spacedaroundthecircumference.Assortedvisualcuesareprovidedbothinsideontheencirclingwallandoutsideontheenshroudingcurtainofthearena,whichserveasvisuallandmarks.Arefugeboxcontainingbeddingmaterialispositionedunderneathasingleescapeholerandomlyselected,oncepersubjectacrossentiretestingperiod.Noxiousauditoryhigh-speedfan15cmabovetheplatformandvisualtwo150-wattoodlightslocated76cmabovetheplatformstimuliareusedtoelicitescapebehaviorduringthesingle,ve-minute64

PAGE 75

dailytrialforeightconsecutivetestingdays.Themouseisplacedinthecenterofthearenaandthenoxiousstimuliareactivatedatthebeginningofeachtrial,andthenumberoferrorsi.e.,headpokesintonon-escapeholesandescapelatencytimearemeasured.ThenumberoferrorsCPE-FinandescapelatencytimeCPL-Finarerecordedforthenaldayoftesting,aswellastherespectiveaveragesacrossalldaysCPE-AvgandCPL-Avg.Theplatformsurfaceisthoroughlycleanedwithdeodorizingdisinfectanttoremoveolfactorycuesduetostress-inducedurinationand/ordefecation.ThisparadigmisarenementoftheBarnesmazedesignBarnes,1979forrats.Thistaskprimarilyevaluatesspatialreferencememory,butalsoincludesgeneralanxietyandsensorimotorcomponentse.g.,Leightyetal.,2004.2.2.5PlatformRecognitionThePlatformRecognitiontaskutilizesthesamequadrant-labeled,100cmdi-ametercircularpoolapparatusastheMorrisWaterMazetask,exceptinthataprominenttargetcmdiametercircularplatformwitha10cmx40cmblacken-signispositioned0.8cmabovethewatersurface.Oneachofthefourtestingdays,theanimalisreleasedagainstthewallofthesamequadrantforeachoffour60sectrials.Theplatformismovedtoadierentquadrantofthepoolforeachtrial.Thetimerequiredforthemousetolocateandmounttheplatformupto60secpertrialisrecorded,followedbya30secintertrialstay"period,duringwhichtheanimalispermittedtoremainatoptheplatform.Ifthemousefailstolocatetheplatformafter60sec,itisgentlyguidedtotheplatformtobeginthestayperiod,andalatencyof60secisrecorded.TheaveragelatencyonthenaltestingdayPR-FinandtheoverallaveragelatencyPR-Avgarecalculatedandrecorded.Thistaskassessesrecognition/identicationmemoryprocesses.65

PAGE 76

2.2.6RadialArmWaterMazeTheapparatusconsistsofa100cmdiagonalcircularpoolcontainingasix-armed0.5cmlongx19cmwidealuminuminsertwitha40cmdiametercentralarea.Themetalinsertextendsfromthepooloorto5cmabovethewaterline,providinganeutralbackgroundforthetestsubjects.Thetaskisdividedintothreeoruptovethree-dayblocks,"foratotalofbetweennineandfteendays.EachdailysessionconsistsoffouracquisitiontrialsT1-T4,followedbya30mindelay,andendingwithadelayed-recallretention"trialT5.Atthebeginningofthesession,asubmergedclearplatformsametargetasinMorrisWaterMazetaskisplacedattheendofthedesignatedgoalarmrandomlychoseneachday,incontrasttotheMorrismazeprocedure,andtheanimalisplacedintooneoftheremainingnon-goalarmsinarandomizedsequence,facingthecentralswimarea,foreachofthefouracquisitiontrials.Themouseisallowedupto60sectolocatetheplatformand,ifsuccessful,remainsontheplatformfora30secintertrialstayperiod.Ifthemouseentersanon-goalarm,itisgentlydrawnbacktothestartingarm.Also,iftheanimalfailstolocatetheplatformafter60sec,itisgentlyledtotheplatformforthestayperiod.Thelatencytolocatethegoalplatformisrecorded,withtrialsinwhichthemousefailstolocatetheplatformrecordedas60sec.Inaddition,thenumberoferrorsentriesintonon-goalarmscommittedduringeachtrialarerecorded.Experimentersalsorecordanybehavioralanomaliese.g.,perseveration,circling,swimmingdiculty.Aftercompletingthefouracquisitiontrials,themouseisremovedfromthepoolandwarmedunderaheatlamp,thenreturnedtoitscagefor30min.FortheretentiontrialT5,themouseisplacedintothesingleremainingunfamiliari.e.,notusedasthestartarminT1-T4non-goalarm,andtheperformancelatencyanderrorsmeasuresarerecorded.Tenaggregatemeasuresarecalculateduponcompletionof66

PAGE 77

thetask:Averagelate-acquisitionperformanceiscalculatedforbothlatenciesanderrorsusingtheT4databothacrossallsessionsRML-T4andRME-T4andforthenalblockonlyRML-FT4andRME-FT4;averageinitial-acquisitionperformanceT1inthenalblockiscalculatedforlatencyanderrorsRML-FT1andRME-FT1;averageretentionperformanceiscalculatedforbothlatencyanderrorsusingtheT5dataacrossallsessionsRML-T5andRME-T5andforthenalblockonlyRML-FT5andRME-FT5.Hence,bothT4andT5measuresrepresentindicesofworkingmemory.ThistaskextendsthestandardMorriswatermaze,whichmeasuresonlyspatialreferencelearningandmemorycapacity,bycouplingthelearning-acquisitioncomponentT1-T4witharecall-retentiontestingphaseT5.Thenalblockoftestingi.e.,lastthreesessionsaremoreindicativeofoverallcognitivefunctioning,whiletheearlierblockslargelyreectprocedurallearningofthisrelativelycomplextask.TheradialarmwatermazeRAWMparadigmisadown-scaledwater-basedadaptationHydeetal.,1998ofearlierdry-landradialmazesforratshaving8or12armsradiatingfromacentralenclosure.Thiscongurationpermittedoneormorefood-orwater-baitedarmstoserveasthegoalse.g.,OltonandSamuelson,1976;Olton,1977;Crusioetal.,1995.Studieswithmulti-armedmazesColeandChappell-Stephenson,2003suggesttheupperlimitofspatialmemoryinratsisbetween24and32locations.2.3NewParadigmsandParallels:TalesofMiceandMenTheidenticationofhomologousneuralsubstratesinhumansandotheranimalspeciese.g.,Spearetal.,1990associatedwithspeciccognitivedomainse.g.,implicitmemory,episodicandepisodic-likememoryrepresentsanimportantad-vanceinneuroscienceresearch,enablinginvestigatorstodevelop,test,andreneanimalmodelsforhumanneurobehavioraldisorders,includingAlzheimer'sdisease.67

PAGE 78

Cross-speciescomparativeneuroanatomicalstudiesinbothnormalandpathologicalconditions,forexample,emphasizetheroleofthehippocampusornon-mammalianhomologousstructuresinspatialmemoryandlearning,aswellasdeclarativemem-oryingenerale.g.,Squire,1992;Eichenbaum,1999;Mavieletal.,2004.Thepersistenceoflocationmemoryinfood-storingbirdsBiegleretal.,2001,spatial-basedradialmazelearninginmiceCrusioandSchwegler,2005,andobject-positionmemoryimpairmentfollowingtemporallobectomywithconcomitanthippocampec-tomyinpatientsNunnetal.,1999aresignicantexamples.Thesendingshavediverseapplicationsbeyondneuroscience;cognitivescienceandlearningtheoryrelyonbothstructuralandfunctionalunderstandingofanimalnervoussystemsandtheirbehavioralmanifestations.Themicroarchitectureandstructuralinterconnec-tivityofthehippocampus,forexample,hasinspiredtheoreticalmodelsofHebbianlearningassociativesynaptogenesisandarticialneuralcomputatione.g.,Kelsoetal.,1986;Fosteretal.,2000whichareusedinadvancedcomputingarchitectureshavingautonomouslearningcapabilitiesarticialintelligence".Specializedcognitivetasksusedtoexplorehumanmemoryandlearningabilities,aswellastodiagnoseclinicalneuropathology,havebeenadaptedforneurobehav-ioralassessmentinanimals,includingmice.Specically,severalformsoflearninge.g.,associative,discriminationaswellasbothlong-termandshort-termmem-orysystemshavebeenexaminedthroughnaturalisticandarticialexperimentalparadigms,oftenveryelaborate,inwhichmultipleparameterse.g.,stimulusdi-mensionandcomplexity,presentationlatencyanddurationaremanipulatedandresponsesrecorded.Indevelopingtheseparallelassessmentmethodologies,however,itisimportanttorecognizethatalthoughhuman-basedtasksforassessinglearningandmemoryoftenrelyuponverbalresponses,whileanimal-basedtasksgenerallyuti-68

PAGE 79

lizemotorresponsese.g.,turndirection,arm-visitlatency,bothresponseclassesrepresentvalidbehavioralindicesformeasurement,comparison,andanalysis.Theserialreactiontimetask,forinstance,involvesmeasuringaccuracyandre-sponselatencyacrosstrialsinasequence-learningproblemNissenandBullemer,1987toassessproceduralimplicitmemoryfunction.Diagnosticimagingstudiesinnormalandtask-impairedindividualsimplicatetheroleofbasalganglia,neocortex,andcerebelluminthistaskKnopmanandNissen,1991;WillinghamandKoroshetz,1993;Pascual-Leoneetal.,1996;Rauchetal.,1997.Theroleofthebasalgangliainimplicitmemoryisunderscoredbylesionstudiesinratswhereinstriatallesions{butnothippocampallesions{resultinproceduralmemoryacquisitionimpairmentinaxed-sequencearm-openingvariantoftheeight-armedradialmazetaskDe-CoteauandKesner,2000.SequencelearningimpairmentisalsoobservedfollowingdorsalcaudatelesionsChristieandDalrymple-Alford,2004inratstrainedin4-,8-,and12-trialarmchoicesequences;performancedecitsarenotobservedinanimalsfollowingdorsalhippocampallesions.ChristieandHersch04adaptedtheappa-ratusformice,andusedmultiplerepetitionsoffour-trialsequenceswithintrainingsessions.Interferenceeectsdecliningperformanceonsubsequentsessionresultedfromreplacingtherepeatingsequencewitharandomsequenceduringasingleses-sionChristieandHersch,2004,similartothepatternobservedinhumanse.g.,NissenandBullemer,1987;ReedandJohnson,1994.Anothermouse-versionoftheserialreactiontaskChoetal.,2007usesanoperantchamberwithnosepokeholespositionedonthewalls;miceconditionedFR-1scheduletorespondthroughnosepokeofilluminatedholesaretrainedtonosepokeaspecicrepeatingsequenceofholes,usingwateraccessasreinforcementforcorrect-sequenceresponses.EvidenceofPavlovianandoperantSkinnerianconditioning,bothexamplesofassociativelearn-ing,iswell-documentedforabroadrangeofanimalspeciese.g.,Spearetal.,1990.69

PAGE 80

Inaddition,non-humananimalsexhibitaremarkablecapacityforlearningcomplexstimuli,bothnaturalandarticial,subjecttospecies-specicsensoryconstraints.VisualdiscriminationexperimentsinpigeonsColumbaliviawhichhavehighly-developedvisualsystems,forexample,indicatethatanimalscanlearnarbitrarycategoriesofnovelstimuli.Theseprocessesrequirepatternrecognition,associativememory,andvisuospatialmemory.Examplesofvisually-presentedstimulusclassessuccessfullydistinguishedbypigeonsinclude:picturesofnaturalobjectse.g.,trees;Vaughan,1988,lightemittingdiodearraypatternsJitsumorietal.,2002,andpaintingsbyMonetandPicassoWatanabeetal.,1995.Similarcomplexlearninghasbeenobservedinratsusingodorrecognition/discriminationtaskse.g.,Kesneretal.,2002,whichalsounderscorehippocampalinvolvementintemporal-sequenceepisodic-likelearning.Event-relatedepisodic-likememoryMorris,2001;HamptonandSchwartz,2004hasbeenexaminedinmiceusingobjectexplorationtasksbaseduponspatiotemporalorderingschemesDereetal.,2005.Thesestudiesdemonstratethatmiceareabletodistinguishnovelfromfamiliarobjectswhat",therelativerecencyofencounterwithanobjectwhen",andthelocationwhereobjectswerepreviouslypresentedwhere"Dereetal.,2005.Additionally,episodic-likemem-orydecitsarestronglycorrelatedwithbeta-amyloidplaqueburdeninAlzheimer'sdouble-transgenicAPPswe/PS1miceSavonenkoetal.,2005,underscoringthediagnosticutilityofepisodicmemorytasksinAlzheimer'sdisease.Workingmemoryprocesses,aswell,havebeenexaminedinbothhumansandanimalsthroughcross-speciesexperimentalparadigms.Forinstance,theHebb-Williams"mazeHebbandWilliams,1946,widely-regardedasthestandardin-strumentforevaluatinganimalspatiallearning,hasbeenimplementedasavirtualenvironment"computersimulationforhumansubjectsShoreetal.,2001whereinthephysicalfeaturesofamazee.g.,walls,alleys,cornersarerenderedwithgraph-70

PAGE 81

icalalgorithms.Virtualmazesrepresentananalogousmaze-learningparadigmforcomparinghumanandanimallearningcapacityunderexperimentalconditions,aswellasforevaluatingneuropathologicalsyndromeswhichmanifestdistinctspatialproblemsolvingimpairment.Forexample,comparisonsbetweentheaverageacqui-sitionperformanceofyounghumansuniversityundergraduatestudentsandmiceC57BL/6Jstrainacrossabatteryoftwelvedistinctmazecongurationsvirtualsimulation,forhumans;actualimplementation,formiceshowedthat,althoughhu-manslearnthetaskmorequickly,thelearningcurvesarestrikinglysimilarShoreetal.,2001.SubsequentcorrelationanalysisamongthesetasksrevealasignicantassociationbetweenmentalrotationandprobetrialperformanceMorrismaze,butnotbetweenotherspatialperformancemeasuresAsturetal.,2004,suggestingdif-ferentialassessmentofspatialproblemsolvingabilitybytheMorrismazeandradialmazeinhumansubjects.Furthermore,humanpatientswithsurgically-inducedme-dialtemporallobelesionsexhibitperformancedecitsinroom-basedanaloguesoftheMorrismazeandradialmazetasksBohbotetal.,2002reminiscentofcognitiveimpairmentobservedinrodentshavingsimilarlesions.ThesendingsareconsistentwiththeresultsofcomprehensivebehavioralassessmentinAlzheimer'stransgenicmice,inwhichcomponentmeasuresobtainedfromspatiallearning/memorytasksMorrismaze,radialarmmaze,platformrecognitionexhibitco-localizationwithincognitivedomains"e.g.,Leightyetal.,2004;Arendashetal.,2007;Leightyetal.,2008,aswellasdiscriminabilitybasedupontransgenicityand/ortherapeutictreatmenteectse.g.,Arendashetal.,2001;ArendashandKing,2002;KingandArendash,2002a;Leighty,2003;Arendashetal.,2004b;Leightyetal.,2004;Jensenetal.,2005;Leightyetal.,2008.Apromisingresearchdirectionforcomparativepsychologyandbehavioralneu-ropathologyinvolvesinterferenceeectsonworkingmemoryandlearning.Interfer-71

PAGE 82

enceisaformofforgetting"whichoccurswhenonerecallprocesscompromisesanotherPostmanandUnderwood,1973.Twotypesofinterference{proactiveandretroactive{aredierentiatedbythetemporaldependencyofforgetting.Inproactiveinterference,earlierlearninginterfereswithsubsequentlearning,aswhenaprofessionaltypistexperiencesdicultylearningtoplaythepiano.Bycontrast,retroactiveinterferenceoccurswhenrecentlearninginterfereswithrecallofearlierlearningas,forexample,whenanindividualhastroublerememberinghispreviousaddressortelephonenumberafterrecentlyrelocatingtoanewcityandlearninganewaddressandphonenumber.Exposuretoacomplex,overstimulatingenvi-ronmente.g.,Toer'sinformationoverload"isacommoncauseofinterferenceinhumansandotheranimals,aswell,andsustainedpressuretoadapttoacceleratingchangeisassociatedwithbothcognitiveimpairmentconfusion,disorientationandphysiologicaldysfunctionanxiety,stresse.g.,Toer,1970.Radialmazelearninginrats,forexample,issubjecttointertrialproactiveinterferencee.g.,Cohenetal.,1996;RobertsandDale,1981whenthedelayintervalbetweensuccessivetrialsisverybrief;theinterferenceeectislikelyduetoerrorsintemporaldiscrimina-tionamongeventsoccurringwithin-vs.between-trialsRobertsandDale,1981.Additionally,interferenceeectsobservedinagedratsinvisualdiscriminationtaskse.g.,Winocur,1984resembleperformancedecitsinhippocampally-lesionedyounganimals,suggestingtheutilityofinterferencetestingforneuropathologicalassess-ment.Inhumans,interferenceeectshavebeenexaminedwithobjectrecognitionandnamingtasksinwhichsubjectsareaskedtorecalltheitemsinagroupoffamil-iarobjectspresentedeitherwithorwithoutintervening,unrelateddistractor"itemse.g.,Fuld,1981.TheFuldtaskhasbeenadaptedbyLoewensteinetal.04,forbothproactiveandretroactiveinterferenceeects,asadiagnosticinstrumentforevaluatingprobableAlzheimer'sdiseaseandmildcognitiveimpairmentinagedindi-72

PAGE 83

viduals.IndividualswithmildAlzheimer'sdiseasecanbedistinguishedfromnormalagedadultswithremarkableaccuracy4.6%sensitivityand96.2%specicityusingmeasuresofbothproactiveandretroactiveinterferenceLoewensteinetal.,2004.Theinterferencetestingparadigmhasbeenadaptedformicebysubstitutingaspa-tiallearningtaskradialarmwatermazeforthehuman-basedverbalsemanticlearningtaskdiscussedinChapter7.73

PAGE 84

CHAPTER3BEHAVIORALDATAANALYSIS3.1OnBehavioralClassicationBehavioralevaluationisonlytheinformation-gatheringphaseofneurobehavioralinvestigation,andmustnecessarilybefollowedbyananalyticprocessforidentifying,quantifying,andinterpretingmeaningfulpatternsandtrendsinthedataset.Thereareseveralimportantreasonsforcollectingandstudyingthesedata,including:be-havioralphenotypingcharacterizinggroups,e.g.,bytransgenicityorstrain,intermsofbehavior;therapeuticmonitoringe.g.,toevaluatetheeectsofongoingpharma-cologictreatments,eithercross-sectionallybygrouporlongitudinallybyindividual,toobservetrendsovertime;andtreatment-outcomeassessmente.g.,lookingattheend-resultofatherapeuticinterventionormanipulation.Alloftheseproceduresin-volveclassication,theassignmentofindividualstogroupsonthebasisofmeasuredvariables.Unfortunately,choosinganappropriateclassiermethodologydependsuponthedata,thehypothesesunderinvestigation,andtheavailablecomputingre-sources,amongotherconsiderations;justasthereisnofreelunch,"thereisnouniversally-superiorclassicationalgorithmWolpertandMacready,1997.Severalbroadcategoriesofclassiersexist,however,andtherearehybridandensemblecom-binationsofmethodswhichmayprovideevenbetterperformancefortheirparticularapplicatione.g.,Sigutetal.,2007.Expertsystems,forinstance,encapsulatetheknowledgeanddecision-makingcapabilitiesofhumanexpertswithinspecicprob-74

PAGE 85

lemdomainsGiarratanoandRiley,2004,suchasdiagnosingbacterialinfectionsbasedonpatientsymptomsandmicrobiologylabresultsMYCIN;Shortlie1974,1976.Behavioralassessment,inpractice,generallyoccursunderconditionsandcon-straintswhicharefarfromwhatistheoreticallyideal.Theeconomicandmanage-mentaspectsofanimalmaintenanceandtesting,forinstance,oftenimposeseverelimitationsonfeasibility.Unfortunately,thediscrepancybetweentheactualandthedesirableinmanysituationscannotberemedied,onlyaccomodated.Considertheproblemofadequatesamplesizee.g.,RaudysandJain,1991:asamplemustbesucientlylargetoberepresentativeofatheoretically-immensepopulationofcan-didateindividualsbut,moreover,thesamplemustbemanageable;hence,limitationengenderscompromise.Balancingacrossgroupsisgenerallymoreimportant,how-ever;ChandrasekaranandJain1979,forexample,notedperformancedegradationinparametricstatisticalclassiersresultingfromunequalsamplesizes.Addition-ally,thepracticallimitsimposedincertainbehavioraltestse.g.,radialarmwatermazetrialshavea60sectime-limit,whilebaseduponestablishedperformancenormse.g.,typical"responseoftestsubjects,constitutedatacensorshipandrequirespecialanalyticconsideration.Gehan65andBreslow1970,forexam-ple,advocatedistribution-freenonparametricstatisticaltestsforcomparinggroupssubjecttoarbitraryright-censorship,basedonWilcoxontwo-groupandKruskal-Wallismultiple-groupmethods,respectively.AmodiedGehan-BreslowanalysiswasusedbyAlkon74instudiesofphotoperiod-mediatedassociativelearninginthenudibranchmolluscHermissendacrassicornis,subjecttotime-limitsi.e.,right-censored.Neurobehavioralresearchers,however,generallyconsidertime-limitasaproceduralissueandutilizestandardMANOVAorKruskal-Wallismethodsforcomparinggroupse.g.,Kingetal.,1999;KingandArendash,2002a.75

PAGE 86

Expediencyinresearchoftenhasaprice;practitionerswhotakeshort-cutswithexperimentaldesignanddataanalysis,regardlessofmotivation,mustproceedwithextremecaution.Inappropriateselectionofrepeated-measuresANOVAinsteadofnonlinearmixed-eectsmodelsforevaluatingpharmacologicdose-reponseeectsinratsKristensenandHansen,2004,forinstance,disregardsthenonlinearresponsecharacteristicstypicalofphysiologicalphenomenae.g.,Peeketal.,2002;KristensenandHansen,2004.Additionally,violationsoftheunderlyingassumptionsofsta-tisticaltestscaninterferewithinterpretabilityor,atworst,compromisethevalidityoftheresultsMontgomery,1953;Boneau,1960.Inpractice,however,thefun-damentalassumptionsofanalysisofvariance{independenceofcases,normalityofdistributions,homoscedasticityhomogeneityofgroupvariances,andsphericityhomogeneityofpaired-dierencevariances{cansometimesbeviolatedwithoutseriousconsequences.Aseriesofpreliminarytestsandcorrectionmethodscanbeusedtodetectandamelioratepotentialviolations,forexample:Kolmogorov-Smirnovtestfornormality;Lilliefors,1968,Bartletttestforhomoscedasticity;Bartlett,1937,1947,Greenhouse-Geissercorrectionforsphericity,andBonfer-ronicorrectionformultiplecomparisons.Inaddition,oneofthebest,albeitleastsophisticated,techniquesforidentifyingpotentialproblemsistoinspectanappro-priategraphicaldepictionofthedata,suchasahistogramorscatterplot;cursoryexaminationoftherawdata,inasuitableformat,isoftenthemostecientmannerfordetectinganomaliesandoutliers.Datatransformationse.g.,BoxandCox,1964;ScottandWild,1991;BlandandAltman,1996aaresometimesperformedpriortoanalysis,particularlywhenparametricstatisticaltestsareplanned,tocorrectissueswhichwouldotherwisecomplicateorinvalidatetheresults.Transformingdatathroughlineararithmeticscalinge.g.,dividingeachvaluewithinacolumnbythevarianceofthesamecol-76

PAGE 87

umn,toproduceunitvariance,forinstance,changesthemeanandvarianceofthecolumnrelativetotheoriginal,butleavesthecorrelationcoecientPearson'sproductmomentcorrelationcoecientbetweentransformedcolumnsunchangedButler,1982.Thisintermeasurecorrelation-preservingtransformationisparticu-larlyusefulforstandardizingdataforneuralnetworks,whichoftenconvergesoonerwith"zero-mean,unit-variance"dataHaykin,1999.Logarithmictransformse.g.,Montgomery,1953;Martoneetal.,1984;Traneletal.,1994;BlandandAltman,1996bareusefulforcorrectingheterogeneityofvariance,andoftenusedforright-skewedpositive-valueddatalargevalues,e.g.,responselatenciesandtoconvertmultiplicativemaineectsintoadditiveeects;thisapproachwillnotworkwithnegativevalues,andverysmallvalues<1shouldbepre-multipliedbyalargeconstant.Squareroottransformsareoftenusedwithcountdatae.g.,bacterialcoloniesonplates,andwhenthevarianceisproportionaltothemean;thesquareroottransformisalsousefulfornormalizingPoisson-distributeddata.Whendataareexpressedasproportionspercentagesand,inparticular,whenthesedataarebinomial,thearcsineorarcsinesquare-roottransformisoftensuggested.Resultscanbeback-transformedusingtheinverseoperation,forsubsequentinterpretationandreporting.Additionally,rank-transformationofdatai.e.,replaceeachvaluewithitsrankfollowedbyANOVAissometimesusedwhentheoriginaldataneithersatisfytheassumptionsforstandardANOVA,normeetthecriteriafornonparamet-ricanalysisConoverandIman,1981,althoughthistransformationcancompromisetestingforeectinteractionsSeamanetal.,1994.Aclassierrepresentsamathematicalmodelofempiricaldatameasuredvari-ableswhichisintendedtosplitpartitionthedataintotwoormoregroups.However,partitioningonesetoflinearlyseparablemultidimensionaldatafromanotherrequiresaseparatorwiththesamenumberofdimensionse.g.,inonedi-77

PAGE 88

mension,apoint;intwodimensions,aline;inthreedimensions,aplane;infourormoredimensions,ahyperplane;thisisthegeometricbasisoflinearclassication.Manysetsofmultidimensionaldataarenotlinearlyseparable,butcanbepartitionedusingnonlinearclassiers,suchasneuralnetworks.Hence,foranygivendataset,therearelikelytobeseveralalternativeclassierarchitecturestobecompared.Acommonmethodforevaluatingthesemodelsinvolvespartitioningthedatasetintothreesubsetsbyrandomassignmentofindividualcases:trainingdatawhichwillbeusedtoconstructtheinitialclassiermodel,testingdatausedtorenethemodeltoimprovegeneralizability,andevaluationdataforsimulatingreal-worldapplicationofthemodelusingunfamiliardata.Therelativeproportionsofthesesubsetsvarieswiththeapplicationandthepractitioner.However,insitutationswheredataarescarceorwhencriticalinstancesmustnecessarilybeincluded,cross-validationtechniquesareusede.g.,Stone,1977;Goutte,1997.Threefoldcrossvalidation,forexample,involvessplittingthedatasetintothreeparts{twotrainingsetsandonetestingset{andrepeatingtheevaluationprotocolthreetimes,thusal-lowingallcombinationsoftestingandtrainingsubsetstobeassignedandevaluated.Leave-one-outk-foldcross-validationjackkning"isamorerigorousapproachcomputationally-intense,inwhichamodelisconstructedusingallbutasinglecase,whichissubsequentlyusedtoevaluatethemodel,andtheconstruction-evaluationprocessrepeatedk-timesi.e.,foreachofthecasesinthedataset.Anotherap-proach,calledbootstrapping,"generatestrainingandtestingsetsusingsamplingwithreplacementfromthedatasete.g.,Efron,1983.Jackkning,however,tendstoproducerelativelyunbiasedresults,comparedwithresamplingtechniques,partic-ularlyforsmallsamplesizesLanceetal.,2000.Comparingtheperformanceofmultipleclassiersrequiresseveralparameterscal-culatedfromtheoutputgeneratedbyeachmodele.g.,AltmanandBland,1994.A78

PAGE 89

classicationmatrix,orconfusiontable,"forexample,oftenprovidesusefulperfor-mancedata;therowsofthematrixdepictactualtruecategories,whilethecolumnsdepicttheclassier-predictedcategories.Thetrue-positive"TPandtrue-negative"TNtotalsrepresent,respectively,thenumberofcorrectly-categorizedpositive"e.g.,transgenic,treatedandnegative"e.g.,non-transgenic,control-groupcaseinstances.Thefalse-negative"FNtotalindicateshowmanytrue-positiveindividualsaremisclassiedasnegative"and,correspondingly,thefalse-positive"FPtotalshowshowmanytrue-negativeindividualsarecategorizedaspositive."Theoverallsuccessrate,oraccuracy,oftheclassieristheratioofcorrectly-classiedcasestothetotalnumberofcases.Mathematically,Accuracy=TP+TN TP+TN+FP+FNThesensitivityistheproportionoftrue-positivecaseswhichareidentiedaspositive.Indiagnosticterms,sensitivityistheprobabilitythatthetestispositivegiventhatthepatienthasthedisease.Expressedas,Sensitivity=TP TP+FNBycontrast,thespecicityistheproportionoftrue-negativecasescorrectlyrec-ognizedasnegativeor,clinically,theprobabilitythatthetestisnegativegiventhatthepatientdoesnothavethedisease.Computedas,Specificity=TN TN+FP79

PAGE 90

Hence,inastudycomparingspatialreferencememoryoftransgenicTg,treat-mentgroupandnon-transgenicNT,controlgroupanimals:thehighertheaccu-racy,themoreTgandNTanimalscorrectlyassignedintotheirrespectivegroups;thehigherthesensitivity,themoreTganimalsrecognizedasTgandnotmisclassiedasNT;thehigherthespecicity,themoreNTanimalsidentiedasNTandnotmis-classiedasTg.IntermsofhypothesistestingNeymanandPearson,1933/1967,false-positiveoutcomesrepresenttypeIerrorsi.e.,therejectionofatruenullhy-pothesis,whilefalse-negativeoutcomesrepresenttypeIIerrorsi.e.,failuretorejectafalsenullhypothesis.Classierperformanceissometimesevaluatedusingare-ceiveroperatingcharacteristicROCcurveHanleyandMcNeil,1982;ZweigandCampbell,1993;Obuchowski,2003;Laskoetal.,2005;Fawcett,2006,whichde-pictssensitivity,ortruepositiverate,"alongtheordinateagainst-specicity,orfalsepositiverate,"alongtheabcissa,fordierentvaluesofadiscriminativeparametere.g.,thecuto-valueforaclassier,testresultcriterion.Thismethodisusefulforne-tuningindividualclassiersandfortestingcomponentclassierstobecombinedintoensemblemodelsSwets,1988;Swetsetal.,2000.Anadditionalstatistic,Kappa,representingtheproportionofsuccessfully-classiedcasescorrectedforchanceperformance,issometimesreported,aswellCohen,1960;Fleiss,1971;VieraandGarrett,2005forcomparingtheperformanceofmultipleindependentclassiers.3.2StatisticalAnalysisConventionalanalyticprotocolsforbehavioralexperimentsgenerallyinvolveoneormoreofthefollowing:examiningrelationshipsbetweenmanipulationse.g.,levelsofanindependentvariableandobservable/measurablephenomenae.g.,dependentvariables,forexample,dose-responsepatternsortreatmenteects;identifyingas-80

PAGE 91

sociationsbetween/amongdependentmeasures,forexample,correlationbetweenbehavioraldecitsandhistopathologicalmarkers;characterizingclinicalsyndromesintermsofbehavioralresponsepatterns,forexample,motordecitsinParkinson-ism;classifyingorcategorizingsubjectsintermsofbehavioralresponsepatterns,forexample,distinguishingbetweennormal-agedandAlzheimer'spatientsusingcog-nitiveassessmentinventories.MultivariatestatisticaltechniquesTabachnickandFidell,2001arecurrentlythestandardapproachintheseinvestigationsforidentify-ing,testing,measuring,andreportingexperimentalgroupdierencese.g.,analysisofvariance,orANOVA,associatione.g.,correlation,latentpatternsindatasetse.g.,factoranalysis,andclassicatione.g.,discriminantanalysis.Theavail-abilityofpowerfulcomputingtechnologyandready-to-runcanned"softwaree.g.,SYSTATTM,StatisticaTM,RTMenablesanyresearchertoperformcomplexanalyseseasilyandeciently.Thissectionpresentsanddiscussesthreeofthemostimportantstatisticalmethod-ologiesforneurobehavioralresearch:correlationanalysis,factoranalysis,anddis-criminantanalysis.3.2.1CorrelationAnalysisCorrelationanalysismeasuresthelinearassociationbetweentwoormorevari-ables.Thisrelationshipisexpressedasthecorrelationcoecient"Pearson'sproduct-momentcorrelationcoecient,r,whichisareal-valuednumberbetween-1.0and+1.0.Mathematically,thisvalueiscalculatedbydividingthecovarianceoftwovariablesbytheproductoftheirstandarddeviations.Themagnitudei.e.,absolutevaluerepresentsthestrengthoftheassociation,withincreasingvalueindicatinggreaterassociationbetweentwomeasures.Negativevaluesdenotepairsofvari-ablewhichchangeinoppositedirections,i.e.,onevariableincreasesastheother81

PAGE 92

decreases;positivevaluesindicatechangeinthesamedirection,i.e.,bothvariableseitherincreaseordecrease.Independencebetweentwovariablesisrepresentedwithacorrelationcoecientofzero.Itisimportanttoemphasizethatcorrelationonlyre-ectslinearassociationbetweentwomeasuredvariables;correlationdoesnotimplyacausalrelationshipbetweenthevariables.Inaddition,correlationswhicharestatis-ticallysignicantarenotnecessarilymeaningful,inapracticalsense.Finally,whenmultiple,simultaneoustestsofcorrelationareperformedonadataset,theresultantsignicancevaluesp-valuesshouldbeadjustede.g.,Bonferroni'scorrectionpriortointerpretation.Correlationanalysesarefrequentlyusedtoexplorecomplexrelationshipsamongbehavioraland/orpathologicmeasuresinexperimentalparadigms.Forexample,ex-tensiveinter-taskandintra-taskcorrelationswereidentiedinamultimetricbehav-ioralassessmentbatteryArendashandKing,2002;Leightyetal.,2004.Signicantintertaskcorrelationswerefoundbetweensensorimotortaskse.g.,BalanceBeamandStringAgility,cognitivetaskse.g.,Y-MazeandCircularPlatform,aswellasbetweensensorimotorandcognitivetaskse.g.,OpenFieldandCircularPlatformintheage-matchednontransgenic-controlandTg2576Alzheimer'stransgenicmiceArendashandKing,2002.Similarly,signicantintertaskcorrelationsbetweensen-sorimotortasksOpenFieldandY-MazeEntriesandcognitivetasksMorrisWaterMaze,CircularPlatform,PlatformRecognition,andRAWM,aswellasbetweensen-sorimotorandcognitivetasksOpenFieldandRAWM,BalanceBeamandRAWMwereobservedinfourAlzheimer'stransgenicmouselinesLeightyetal.,2004.Inaddition,performancemeasuresshowedsignicantintrataskcorrelationwithinboththeCircularPlatformandMorrisWaterMazetasksArendashandKing,2002inTg2576mice.Likewise,signicantintrataskcorrelationwasfoundintheCircularPlatform,MorrisWaterMaze,PlatformRecognition,andRAWMtaskscompletedby82

PAGE 93

fourAlzheimer'stransgenicmouselinesLeightyetal.,2004.Subsequenthistolog-icalexaminationofbraintissuesfromTg2576transgenicmiceusingsynaptophysinimmunostainingrevealedthatincreasedhippocampalstainingisassociatedwithbothsensorimotorBalanceBeamandcognitiveMorrisWaterMazeperformancedecitsintheTg2576miceKingandArendash,2002b.Inaddition,theY-MazeentriesmeasurewassignicantlynegativelycorrelatedwithCongophiliccompactbeta-amyloidlevelsinbothcortexandhippocampusofAlzheimer'stransgenicmiceLeightyetal.,2004;PlatformRecognitionandRAWMT4L,T5Lmeasureswerealsosignicantlypositivelycorrelatedwithbothcorticalandhippocampalcompactbeta-amyloid.Totalcorticalandhippocampalbeta-amyloidE10immunostaininglevelsweresignicantlycorrelatedwithBalanceBeamnegatively,andMorrisWa-terMazeacquisition,PlatformRecognition,andRAWMallpositivelyLeightyetal.,2004.SignicantpositivecorrelationbetweenalternativepsychometricbatteriesFil-lenbaumetal.,1987usedforclinicaldiagnosisofAlzheimer'sdiseasereectsthesimilarityofcognitiveabilitiesassessedbytheseinstruments.Indeed,multitask-multimetriccognitiveassessmenttypicallyresultsinpositivemanifoldcorrelationstructure,i.e.,subjectstypicallyretainrank-orderingacrossproblem-solvingtasks,whichhasbeentakenasevidenceofasingleunderlyingcognitiveconstructsinceSpearman's904elucidationofageneralintelligence"factor.Additionally,post-mortemstudiesinAlzheimer'spatientsshowedsignicantcorrelationbetweenre-gionalfrontalcortex,hippocampuscholinergicactivityandcognitiveperformanceMMSEandMRDStestsmeasuredwithinayearofdeathPappasetal.,2000,underscoringthediagnosticutilityofbehavioralpsychometricassessment.83

PAGE 94

3.2.2FactorAnalysisFactoranalysisisamultivariatedatareductiontechniqueforidentifyingunder-lyingassociationsamongmeasuredvariablesandusingthesepatternstogenerateasmallerensembleofcompositevariablescalled"factors".Exploratoryfactoranaly-sis,themostcommonmethodology,attemptstoextractthelatentassociativestruc-tureamongvariableswithoutusingpriortheory,allowingthepractitionertointer-prettheresultingfactorloadingsmatrix.Conrmatoryfactoranalysis,bycontrast,determineswhetherthenumberandcompositionoffactorsextractedfromagivendatasetconformtoapre-establishedusuallytheoreticalexpectation.Theextrac-tion"offactorsisaniterativeoptimizationprocedureinwhichvariance-maximizinglinearcombinationsofthemeasuredvariablesaregenerated,theresultingvarianceissubtractedandasecondlinearcombinationgenerated,andsoforth;thisapproachiscalledprincipalcomponentsanalysis,"andresultsinorthogonali.e.,uncorrelatedfactors.Alternativeextractionalgorithms,suchasprincipalfactoranalysis,"canbeusedtodeterminetheminimumnumberoffactorswhichaccountforthecommonvarianceofadatasete.g.,VelicerandJackson,1990.Thefactorloadingsmatrixcomponentloadings,inprincipalcomponentsanalysisrepresentsthecorrelationco-ecientsbetweenfactorscolumnsandcomponentmeasurevariablesrows;hence,thesevaluesreecttheamountofeachcomponent'svarianceexplainedbyeachfac-toronwhichitloads.Frequently,whentheinitialcomponentloadingpatternsarediculttointuit,additionalmathematicalprocessingrotation"isperformedtoimproveinterpretation;VarimaxrotationKaiser,1958,forexample,generallypro-ducesafactormatrixinwhicheachcomponentvariableisassociatedwithasinglefactorinsteadofloadingonseveralfactors.84

PAGE 95

Factorsextractedarenotnecessarilyfactorsinterpreted;thepractitionerisul-timatelyresponsibleformakingsenseofeachfactorbased,inpart,uponexpecta-tionandpastexperiencee.g.,Fabrigaretal.,1999.Researchersuseacollectionofheuristicsforinterpretingfactoranalyticresults,supportedbytheoretical,em-pirical,andsometimesanecdotalevidenceZwickandVelicer,1986;Lanceetal.,2006.Theeigenvalueorcharacteristicroot"ofeachfactorreectsthepropor-tionofoverallvarianceofallmeasuredvariableswhichisaccountedforbythatfactor;eigenvaluesgreaterthanunity.0,forexample,areofteninterpretedassignicantGuttman,1954;Kaiser,1960.Factorshavingfewerthanthreehigh-magnitudecomponentloadings,forinstance,shouldnotbeinterpretedVelicerandFava,1998.Inaddition,contrarytocommonpractice,largersamplesizesaloneareinsucienttoimprovefactorpatternresolutionGuadagnoliandVelicer,1988;MacCallumetal.,1999,althoughthenumberofcasessamplesmustalwaysexceedthenumberoffactors.However,MonteCarlosimulationshaveshownthatfactorstructureandthemagnitudesofloadingscaninuencefactorscores,particularlyinthecaseofsmallsamplesGrice,2001.Finally,factorstructurecanbealteredsometimesprofoundlybytheinclusionofadditionalbehavioralmeasureswithinasingletask.RodgersandJohnson1995,forexample,identiedtwofactorsanx-iety"andlocomotoractivity"usingthestandardspatiotemporalmeasuresi.e.,frequencyanddurationofopen-andclosed-armentriesintheelevatedplus-mazetask.However,athirdfactordecisionmaking"emergedwhenthedurationoftimespentinthemazecenterwasincludedintheanalysis.Indeed,theseexamplesunderscorethepopularmaximthatinterpretingfactoranalyticresultsisasmuchanartasascience.Therstpracticalapplicationoffactoranalysisinvolvedintelligencetestingdur-ingtheearly20thcentury,whenCharlesSpearman1904postulatedacommon85

PAGE 96

underlyingtrait"forhumanintelligencetoexplainobservedindividualcorrelationsbetweencognitivetasks.Indeed,asingleconstructofgeneralcognitiveabilitymayexplaincompetenceacrossabroadrangeofareasacademic,vocational,etc.Kun-celetal.,2004.Manybehavioralresearchers,however,havequestionedtheva-lidityofasingle-factorcognitivetheoryandsubsequentlydevelopedmorecomplexmodelscomprisedofmultipledistinctfactors,eachrepresentinganuniquefacetofcognition.Thurstone938,forinstance,identiedsevenprimarymentalabili-ties"i.e.,associativememory,numberfacility,perceptualspeed,reasoning,spatialvisualization,verbalcomprehension,andworduencybyusingmultiplementalas-sessmentinventories.Humanbehavioralandpsychologicalcharacteristicsarealsoidentiedthroughfactoranalysis.Forexample,fourfactorswereextractedfromacomprehensivepsychologicalassessmentofAlzheimer'soutpatientsN=435,ac-countingfor57%oftotalvarianceMirakuretal.,2004:aect"e.g.,agitation,anxiety,depression;physicalbehavior"e.g.,apathy,disturbancesinappetiteandsleep;psychosis"e.g.,delusionsandhallucinations;and,hypomania"e.g.,dis-inhibition,euphoria.Factoranalysisisusedtovalidatepsychometricinventories,aswell.Forexample,factoranalyticstudiessupporttheve-factorstructureoftheMMSE.Alarge,multisitesampleofolderadultsN=8556;50-80y/o,forex-ample,returnedanobliqueve-factorsolutionJonesandGallo,2000.Thevefactorsare:Concentrationserialsubtraction,backwardword-spelling;LanguageandPraxisthree-stepcommand,read-followinstructions,sentencewriting,poly-goncopying,objectnaming;Orientationtime-andplace-orientation;Memorydelayed-recall;and,Attentionimmediatewordrepetition.Additionally,Ba~nosandFranklin002administeredtheMMSEwithinamoreheterogeneouspopula-tionN=339;18-87y/o;psychiatricpatientsandalsofoundanobliqueve-factorsolution:Orientation,Attention-WorkingMemory,Comprehension-Praxis,Naming,86

PAGE 97

andVerbalRecall.Factoranalysisalsorevealslifespan-developmentaltrendsinpsy-chometricmeasures.Forexample,thetendencyforhumanintelligencetobecomeincreasinglyundierentiatedwithincreasingagee.g.,StrickerandRock,1987;fac-toranalysisofverbal,quantitative,andanalyticalitemsoftheGraduateRecordExaminationTMshowsthat,althoughcognitivefactorsremainlargelydistinct,theintercorrelationofthefactorsincreaseswithage.Manifoldpositivecorrelationacrosscognitivetaskse.g.,Morriswatermaze,Hebb-Williamsmaze,andwaterplus-mazehasbeenfoundinbothratse.g.,Thorn-dike,1935andmiceMatzeletal.,2003;Galsworthyetal.,2005,aswell;unrotatedprincipalcomponentsanalysisreturnsarstfactorwhichaccountsforbetween38and61percentoftaskvariance.Additionally,thestrengthofcorrelationbetweentasksin-creaseswithincreasingproblem-solvingcomplexityThorndike,1935.Mouse-basedstudiesKolataetal.,2005alsoindicatesignicantcovarianceofgeneralcognitiveabilitywithshort-termworkingmemory,butnotwithlong-termretentionmem-ory,suggestingtheexistenceofseparatememorydomainscf.Kingetal.,1999;Leightyetal.,2004,asobservedinhumans.Theprimaryfactorinmice,therefore,mayrepresentanoverallcognitiveperformancemetricPlomin,2001;Galsworthyetal.,2005whichparallelsthegeneralcognitiveabilityconstructinhumans.Fac-toranalysishasbeenusedfordimensionalreductionofmultimetricperformanceassessmentsinmice;thethirteenornineteenperformancemeasuresobtainedfromacomprehensivebehavioralassessmentbattery,forexample,canbereducedtotwoorthreerecognizablefactorse.g.,workingmemory,"locomotionandexploratorybehavior";Leightyetal.,2004.Thesefactorscanbeusedsubsequentlytochar-acterizegroupdierencesinbehavior,forexample,sexdierencesintheanxiety"factorextractedfromelevatedplusmazedataofratsFernandesetal.,1999.Thecomplexinterplaybetweencognitiveprocesses,suggestiveofunderlyingneurophysio-87

PAGE 98

logicalconnectivity,isalsoexploredthroughfactoranalysis.Theinterdependenceofhumanworkingmemorye.g.,forwardorreversedigitspanandcomplexcognitivetaskse.g.,arithmeticprocessing,counting,forexample,isrevealedthroughsharedcomponentloadingofpsychometricmeasurese.g.,Miyakeetal.,2001.Similarly,theinuenceofnoncognitiveelementsofmousebehavioraltasks,suchasspecies-specicswimmingpatternsintheMorriswatermaze,canbeidentiedthroughfac-torloadingpatternsWolferetal.,1998.Sensorimotorcomponentscanintroducevariability,analogoustoelectronicnoise"Wolferetal.,1998,obscuringthesubtlecognitivemanifestationsofgenotypicdierences.Furthermore,becausetherelativecontributionsofeachcomponentmeasureofallfactorsarecalculatedcalledthefactorstructure,factoranalysesareusefultoolsfordevelopingparsimoniousmodelsofcomplexbehavioralphenomena.Inaddition,factorscores"generatedforeachindividualsubjectfromtheircorrespondingperformancemeasures,forinstancecanberegressedontoothervariablese.g.,pathologicmeasuresPappasetal.,2000.Factoranalyseshavealsobeenusedwithanineteen-measurecomprehensivebe-havioralassessmentbatteryinAlzheimer'stransgenicmice,bothwithandwithoutpathologicmeasurestotalandcompactbeta-amyloidincortexandhippocampus,toexplorepatternsofassociationamongmeasurese.g.,Leightyetal.,2004.Mea-suresfromtheRAWM,PlatformRecognition,andCircularPlatformlatencytaskscomprisetherstfactor,accountingforover32%ofvariance.BehavioralmeasuresfromtheMorrisWaterMaze,Y-Mazeentries,andCircularPlatformerrorsloadedindependentlyasthenextthreefactors,eachaccountingforabout10%ofoverallvari-ance.Whenincluded,allfourpathologicmeasuresloadedwiththeRAWM,PlatformRecognition,andCircularPlatformlatencymeasuresontherstfactor,increasingthevariancecomponenttoapproximately35%.Thesendingsunderscoreboththediagnosticutilityofthethreewater-basedtasksMorrisMaze,PlatformRecognition,88

PAGE 99

andRAWMwhichshareinterrelatedcognitivedomains,andthestrongassociationbetweencognitiveimpairmentreectedasperformancedecitsandneuropathol-ogycorticalandhippocampalbeta-amyloiddisplayedinAlzheimer'stransgenicmiceLeightyetal.,2004.Furthermore,exploratoryfactoranalysesofbehavioralassessmentscompletedatdierenttime-pointsmayprovideimportantinsightsonpatternsofcognitivedevelopmentand/orprogressiveimpairment.Forexample,factor-structurechangeswereobservedinAPP/PS1transgenicmicebetween4.5-6monthsand15-16.5monthsofageJensenetal.,2005.Atthelatertime-point,theprimaryfactorwascomprisedofmeasuresfromtheRAWM,MorrisWaterMaze,andPlatformRecognitiontasks;attheearliertime-point,onlythersttwotaskswererepresented.Inaddition,vedistinctsignicantfactorswereidentiedatthelatertime-point,whileonlythreefactorswerereturnedattheearliertime-point.Theseresultssuggestagrowingdissociationamongcognitivedomainswithincreasingage,perhapsreectingsomeprogressivedecouplingofcognitiveprocesses,concomitantwithadvancingneuropathology.3.2.3DiscriminantAnalysisDiscriminantanalysisisamultivariatestatisticaltechniqueforidentifyingessen-tialvariableswhichreliablydistinguishbetweenoramongpre-denedgroupsandpredictinggroupmembershipforunclassiedindividualsonthebasisofthevariables.Theproceduregeneratesamathematicalmodel,consistingofoneormorefunctionstypicallylinearwhichpartitionindividualsintogroupsusingacombinationofthemeasuredvariables.Thecomplete"ordirect-entry"variantattemptstoincludeallthemeasuredvariablesinthemodel,whilestepwise"e.g.,stepwise-forwardvariantsiterativelyselectandtesteachmeasureforinclusionbasedonvariance-optimization,exhaustivelyaddingand,possibly,removingmeasures.Stepwise89

PAGE 100

methodsoftengeneratemoreparsimoniousmodelsi.e.,havingfewervariablesandexhibitsuperiorclassicatoryability,relativetocompletediscrimiminantanalyses.Forexample,direct-entrydiscriminantanalysisofacomprehensivebehavioraltaskbatteryconsistingofeither15or32sensorimotorandcognitivemeasuresinAPPswTg2576,Tau,andnon-transgenicmicefailedtodistinguishamongthethreegroupsArendashetal.,2004b,however,astepwise-forwardanalysisachievedsignicantdiscriminabilityforbothsetsofmeasuresp<.0001andp<.0005,respectively.Anineteen-measurecomprehensivebehavioralassessmentwasunabletodistinguishamongfourAlzheimer'stransgenicmouselinesusingcompletediscriminantanaly-sis,althoughastepwise-forwardanalysissignicantlydistinguishedamongthelinesusingonlyfourmeasuresYM-Alt,YM-Ent,PR-Avg,andRM-B3T4LLeightyetal.,2004.InclusionoftotalandCongophiliccorticalandhippocampalbeta-amyloidmeasuresalsoresultedinsignicantdiscriminabilityamongthefourmouselinesus-ingstepwise-forwardanalysis,generatingafourpredictor-variablemodelYM-Alt,YM-Ent,RM-B3T4L,andHP-CngaswellLeightyetal.,2004.Hence,thein-clusionofpathologicbiomarkerswithbehavioralperformancemeasuresmayprovidesupplementaldiscriminativeinformationforclassicationanddiagnosticevaluation.Physicalmeasurements,collectedfromindividualstobeclassied,aretypicallyemployedindiscriminantanalysis.Forexample,Fisher's36discriminantfunc-tionanalysisistheclassicdemonstrationofusingmorphometricfeatureslengthandwidthofsepalandpetalstructurestoclassifyindividualsintogroupsthreeIrisspecies:setosa,versicolor,andvirginicausingalinearcombinationofthefourphysicalmeasures.Inadditiontostructuralfeature-basedclassication,function-basedtaxonomiese.g.,behavioralobservationsorcognitiveperformancearealsopossible.BehavioralphenotypingofinbredmousestrainsusingtheSHIRPAtestbatteryRogersetal.,1999,forexample,reliesondiscriminantanalysistoclassify90

PAGE 101

individualanimalsintotheirrespectivestraincategoryonthebasisofsensorimo-torandcognitivetaskperformance.Inaddition,18F-uorodeoxyglucosepositronemissiontomographyFDG-PETwasusedtodistinguishbetweenage-matchedgroupsofADpatientsMMSEcriteriaandhealthysubjects,yieldingsensitivityof90+%andspecicityof95%Habecketal.,2008.Predictinganindividual'sfu-tureperformanceusinghistoricaldatae.g.,graduateschooladmissions,employmentcandidatescreeningisanothercommonapplicationofdiscriminantanalysise.g.,Neely,1977.Discriminantanalysisisalsousefulinclinicaldiagnosis,forgeneratingreduced-lengthcognitiveassessmentbatteries.Forexample,discriminantanalysisofafteen-measureneuropsychologicaltestresultedinasubsetofvemeasuresrepresentingoverallcognitiveabilitywhichaccuratelydistinguishedamongnormal,mildly,andmoderately-to-severelyimpairedgeriatricoutpatientsWhelihanetal.,1997.3.3DataMiningAnalysisDataminingisamulti-stepprocessforexplorationandrecognitionwithincol-lectionsofdataFayyadetal.,1996b.Drivenbytheever-increasingvolumeofdataavailabletoday,thegoalofdataminingistoextract,identifyandanalyzemeaning-fuli.e.,pertinentinformationcontente.g.,consistentpatterns,systematicrulesortrendsinthecontextofextraneousdetail.Someoftheimportantconsiderationsfordataminingoperationsinclude:selectinginformation-appropriaterepresentationschemesdatastructures;e.g.,lists,arrays,orgraphs;designingandimplementingtask-appropriatealgorithmse.g.,depth-rstsearch,feed-forwardbackpropagationoferror,entropyminimization;andevaluatingaccuracye.g.,Aretheanalyticre-sultscorrect?,utilitye.g.,Doestheanalysisrevealnovel,nontrivialknowledge?,andperformanceeciencye.g.,Isthereafasterorless-expensivemethodforob-91

PAGE 102

tainingcomparableresults?.Dataminingtechniqueshaveawiderangeofappli-cationse.g.,Fayyadetal.,1996a;KohandTan,2005,including:nancee.g.,optimalportfolioallocationandvalueprediction,marketinge.g.,identifyingcon-sumergroceryshoppingpatterns,industrye.g.,monitoringplantoperationsandqualitycontrol,andmedicinee.g.,riskmanagementandtreatmentprotocolassess-ment.Furthermore,throughtheintegrationofshareddatabasese.g.,Amarietal.,2002,contemporaryneuroscientistscanexchangebothresearchdataandanalytictoolsandestablishlarge-scalecollaborativenetworksforstudyinghumanbraindis-orders.Thecapacityofdataminingmethodologiesformanagingandcoordinatingimmense,heterogeneouscollectionsofinformationmakestheseresearchinnovationspossible.Dataminingtechniquesusefulforscienticresearchincludebothsupervised-learningtraininginvolveslearningamappingbetweeninputitemsandtheircorre-spondingtargetoutput;e.g.,classiersandunsupervised-learningconstructingamodelusingonlythetrainingitems,withoutaprioristructuralknowledge;e.g.,datacompression,clusteringalgorithmsapproachesforknowledgediscovery,suchasdeci-siontrees,neuralnetworks,supportvectormachines,andothers.Enhancedclassierperformancee.g.,greateraccuracy,speed,and/orgeneralizability,forinstance,isoftenachievedbycombiningseveralsimilarordierentdataminingmethodstoproduceensemblemodelse.g.,BauerandKohavi,1999;ProvostandFawcett,2001.Multiple-classiersystemscanbeusedtopartitioncomplexproblemsfordistributedprocessing,ortoexploittherespectivestrengthsofeachindividualmethodologyDietterich,1997;WolpertandMacready,1997;Kiang,2003.Phillips-Wrenetal.007,forexample,comparedseveraldataminingtechniquesdecisiontrees,neuralnetworks,logisticregressionforpredictingmedicaloncologistvisitsbasedonpatientdatademographics,insurance-eligibility,medicalhistory,andshowedthatneural92

PAGE 103

network-basedclassierperformanceimproveswhenthedataarepre-processedusingdecisiontrees.Thissectiondescribesthreecommondataminingtechniques{decisiontrees,neuralnetworks,andsupportvectormachines{whichhavebeenusedseparatelyfrom,aswellasinconjunctionwith,conventionalstatisticalmethodsforanalyzingbiomedicaldatainbothexperimentalandclinicalcontexts.3.3.1DecisionTreesThisdataminingtechniqueisbasedonagraph-theoreticdepictionofactionsandconsequences,usingthemetaphorofatreewithbranchesandleavestorepresentclassicationrulesandcaseinstances,respectively.Eachbranchcorrespondstoadecisionevent,andthedegreeofarborizationbreadthanddepthofthetreereectsthecomplexityoftheclassier.Ideally,decisioneventscalledsplits"shouldparti-tionthedatasetintogroupscontainingasingledominantclass,andsplittingshouldcontinueuntilauser-establishedcriterionisreachede.g.,allindividualsinagroupareofthesameclass,nosamplecasesremain.Classicationusingdecisiontreesinvolveslearningthroughinductioni.e.,rule-inferencefromexamplesconsistingofattributesandacorrespondingconclusion,informationtheory,andthereductionofentropyuncertainty.Thenumberofbitsdecisionsneededtoclassifyindividualcases{similartotheguessinggameoftwentyquestions"{isrelatedtothedepthofthetree;theobjectiveistochoosesplitsinamannerthatsignicantlyreducesuncertaintyateachstep.Considertheexampleofamulti-taskbehavioralevaluationoftransgenicandnontransgenicmice,forwhichweareprovidedwithatablelistingeachanimal'sperformancemeasureforeachtask,alongwithitsgenotypicidentitytransgenicornontransgenic.Eachbehavioralmeasurerepresentsanattribute,andeachcorrespondinggenotypicidentityrepresentsaconclusion,foreachanimalcase.93

PAGE 104

Thegoalistoconstructatreewithadecisioneventateachbranchpointe.g.,Istheanimal'sstringagilityscoregreaterthanthree?",suchthatbranchesterminateatcorrectconclusions.Aftergeneratingthecompletetree,onecanelucidateoverallrules,forexample,IfananimalaveragesfewerthanveRAWMTrial5errors,andhasabalancebeamlatencyofatleast45sec,thenitisnontransgenic"andintheretention-testphaseoftheMorrismazetask,ifananimalspendslessthan35%ofthetimeinthegoalquadrant,thenitistransgenic."Thistechniquehasseveraladvantages:decisiontreesareeasilygeneratedandreadilyunderstood;bothnominalandcategoricaldatacanbeincludedinthedataset,missingcasesareeasilyhandledasaseparateclass;and,thereisnoblackbox"i.e.,allobservedsplitscanbedescribedintermsofBooleanlogic,asconjunc-tive/disjunctiverules,insteadofcomplexmathematicalformulae.Onesignicantdisadvantageofdecisiontrees,however,istheirsensitivitytosmallvariationsinthedataset.Forexample,theuseofinformation-gainentropyreductioncriteriaateachsplitmeansthatwhentwoormoreattributese.g.,RME-T4andWM-Ret,intheaboveexamplehavenearlythesamecalculatedentropy-reducingeect,theadditionordeletionofasinglecasemaydeterminetheshapeoftheentiretree.Ex-amplesofcomputeralgorithmsfordecisiontreelearningareID3Quinlan,1986,whichisrestrictedtocategoricalvalues,andC4.5Quinlan,1993,whichallowscontinuous-valuedattributes.Decisiontree-basedclassiershavebeenusedforanalyzinglargedatasetsofneu-rophysiologicaldatatoassistmedicaldiagnosis,aswellasinthedevelopmentofbehavior-basedscreeninginventoriesforAlzheimer's.Additionally,decisiontreesareusedinneuropharmacologicalresearchforgeneratingdrugscreeningprotocols.FordetectingepileptiformEEGspikes,decisiontree-basedclassiersJ4.8arecom-parabletovisualanalysisbytrainedhumanexpertsValentietal.,2006.Abi-94

PAGE 105

narydecisiontreewasusedtogenerateaneleven-question,true-falsesurveycalledtheSymptomsofDementiaScreener"Mundtetal.,2000,intendedforusebynonclinicalpersonnelandcaregiverstoscreenforprobable-AD.Theoptimalpro-portionofcorrectly-identiedprobable-ADindividualsi.e.,sensitivitywas90.2%,andtheproportionofcorrectly-identiednon-dementedindividualsi.e.,specicitywas84.6%Mundtetal.,2000;thisinstrumentperformedbetterthanconven-tionalMMSEscoring,whichdemonstrated78.7%sensitivityand92.2%specicity.Finally,pharmacologicresearchonmouse-basedmodelsofdepressioninvolvesmul-tiplebehavioralresponsemeasurese.g.,tailsuspensiontest,forcedswimmingtest;Bourinetal.,2005.However,becausebehavioralresponsesinmiceareinuencedbystrain-dependentdierentialdrugsensitivity,itisoftendiculttointerpretandcompareresultsofexperimentsinwhichdierentstrainsand/ordrugdosagelevelsareexamined.AdecisiontreewasconstructedtoidentifyanappropriateselectionsequenceformousestrainsSwissNMRI,C57B1/6J,orDBA/2andbehavioraltaskstailsuspensionorforcedswimmingtoevaluatecandidateantidepressantcompoundsBourinetal.,2005,basedonmechanismofaction.3.3.2NeuralNetworksArticialneuralnetworksneuralnetworks,"ANNs;Abdi,1994;Haykin,1999areaclassoflearningmethodsinspiredbystructuralandfunctionalfeaturesofbiologicalnervoussystems.Neuralnetworksconsistofalargenumberofinter-connected,independently-operatingcomputingelementswhoseaggregateactivityresultsinclassication,recognition,feature-extraction,andotherhigher-ordercom-putationalcapabilities;theseemergentcollectivecomputationalabilities"Hop-eld,1982arethehallmarkfeatureofcomplexself-organizingsystems,suchasecosystems,cities,andbrains.Therstmathematicalmodelofneuralcomputation95

PAGE 106

basedonthreshold-logicandall-or-noneresponsewasproposedbyMcCullochandPitts943,whoseneuron"receivedexcitatoryandinhibitoryinputinformationthroughsynapses,"performedarithmeticspatialsummation,andproducedbinaryoutput.Later,Rosenblatt58designedatwo-layeredtrainablecomputingmodelforpatternrecognition,calledtheperceptron,"whichcouldonlysolvesimplei.e.,linearly-separableclassicationproblems.Perceptronscannotsolvemorecomplexe.g.,non-linearlyseparableproblemssuchastheexclusive-ORlogicalrelatione.g.,MinskyandPapert,1969.Amultilayered-perceptronarchitecture,however,withatleastasinglehiddenlayerandsucientcomputingelementsbecomesauniver-salfunctionapproximatorwhichcandescribeanycontinuousfunctionKolmogorov,1957;Cybenko,1989;anANNwithtwohiddenlayerscandescribeanyarbitraryfunction.Subsequently,manydierentneuralnetworkcomputingarchitectureshavebeendeveloped,eachutilizingitsownoptimizationparadigme.g.,thermodynamicmetaphorinBoltzmannmachines;Ackleyetal.,1985andinformationrepresen-tationalschemee.g.,self-organizingmaps;Kohonen,1982.Theabilityofneuralnetworkstoconstructnonlinearmappingsbetweenindependentanddependentvari-ablese.g.,betweenbehavioralmeasuresandtreatmentgroupsexempliestheirutilityindiversereal-worldapplicationswheretherelationshipsbetweenoramongmeasuredvariablesareeithercomplexorunclearZhang,2000;Almeida,2002.Hence,neuralnetworksareusefulforconstructingdynamicalmodelsofensemblesystemshavingmultipleinteractingcomponents.Neuralnetworksarerobusthighlyfault-andnoise-tolerant,andparticularlyusefulforhandlingincompleteorinexactdata.TrainingANNs,however,iscomputa-tionallyintenseandrequiresfastprocessingmachinesforecientoperationinreal-worldapplications.Untrainedneuralnetworkscontainnoaprioriknowledgeofthelearningprobleme.g.,theinitialmatrixofinterneuronalconnectionweights96

PAGE 107

containsrandomnumbers,hence,trainedANNssometimesconvergetodierentinternalstableweight-matrixcongurations{eitherspontaneously,orinresponsetosmallinducedperturbationse.g.,inputand/orweightnoise;Pendharkar,2002.Consequently,leave-one-outk-foldcross-validationisrecommendedinadditiontoperformingandcomparingmultipletrainingsessionsCunninghametal.,2000.Furthermore,becauseANNsmakenoassumptionsconcerningtheunderlyingdistri-butionofdatai.e.,nonparametric,theirperformanceisrelativelyimmunetode-parturesfromstatisticalnormalityofdistributionsandlinearityofrelationshipsbetweenvariables.However,becausenospecicationofanexplicitrelationshipbetweenpredictorvariablesandoutcomeisrequiredor{indeed{necessaryforcon-structingandtrainingneuralnetworks,thereiscurrentlynoprescribedapproachforassemblingoptimalnetworktopologiese.g.,numberoflayersorhidden-layerunits.Instead,practitionersoftenrelyonexperience,heuristics,andtrial-and-errorexperimentationforguidanceWalczakandCerpa,1999,e.g.,usingthegeometricmeanofthenumbersofinput-andoutput-layerelementstoassignthenumberofhidden-layerunits.AlthoughANNsarenormallyimplementedassoftwarecomputerprogramsex-ecutedonhigh-performancehardwareplatforms,ageometricmetaphorprovidesausefulmeansofdescription.Onestandardarchitectureforsupervisedlearning,calledafeed-forwardnetwork,iscomprisedofnitesetsofcomputingelementsalsocalledneurons"orcomputingunits"arrangedintodistinctlayers,witheachlayerfully-connectedtotheprecedingandsucceedinglayersi.e.,allelementsarepairwise-connectedbetweenlayersbyweightedlinks.Therstlayer,calledtheinputlayer,"receivesthestimulusdatae.g.,caseinstancesusedfortrainingandtesting,andthenallayer,calledtheoutputlayer,"returnsthenalcomputedresultgeneratedbythenetwork;layersofelements,positionedbetweentheinputandoutputlayers,97

PAGE 108

areknownashiddenlayers."Hence,informationprovidedtotheinputlayerispro-cessedwithintheelementsofthatlayer,thenpassedforwardtothesecondlayerforprocessingwithinelementsofthesecondlayer,andproceedingacrossalllayersuntiltheelementsoftheoutputlayerreceiveandprocesstheinformation.Traininganet-worknecessarilyrequiresstructuralchanges,whicharerepresentedbyadjustmentsoftheconnectionsi.e.,weightsbetweenelements.Themagnitudeoftheweightadjustments,aswellasthechoiceofconnectionstobeadjusted,isdeterminedbycomparingthedesiredoutputofthenetworkrecall,thisissupervisedlearningwithitsactualoutputandapplyinganappropriatelearningalgorithm.Theback-propagationtrainingalgorithmRumelhartetal.,1986,forexample,realizestheprincipleofHebbianlearningHebb,1949,byiterativelyadjustingtheconnectionstrengthsweightsbetweensuccessivelayersofthenetwork,inretrogradefashionstartingfromtheoutputlayer,tominimizethediscrepancyerrorbetweenthede-siredandactualoutputsofthenetwork.TraininginvolvesacompromisebetweenmemorizationandgeneralizationAlmeida,2002,i.e.,betweenacquiringaperfectmappingbetweentrainingexamplesandcorrespondingoutput,andacquiringsu-cientassociationbetweeninputandoutputpatternsforadequatehandlingofnovel,testingexamples;stoppingcriteriae.g.,numberoftrainingepochs,asymptoticerroranalysisareusedtodeterminehowlongtocontinuetraining.Articialneuralnetworksrepresentapromisingnewcomputationaltoolforbothclinicalandexperimentaldataanalysis,aswellasformedicaldecision-makinge.g.,Baxt,1995;Papiketal.,1996.Indeed,diagnosesgeneratedbyneuralnetworksaregenerallycomparable,andsometimessuperior,tothosemadebyhumanexpertse.g.,Milleretal.,1992;TafeitandReibnegger,1999.Forexample,areviewofclinicalandrandomized-controltrialsLisboaandTaktak,2006involvingneuralnetworkadjunctsincancerdiagnosisandprognosisshowedasignicantbenet98

PAGE 109

outof27studiesbyincludingthesemethodswithconventionalmedicalassessment.SimilardiagnosticadvantagesarereportedforthepredictionofmortalityincardiacsurgerypatientsNilssonetal.,2006,dynamicidenticationofgaitpathologiesusingkinematicdataSchollhorn,2004,andevaluationoftreatmentecacyofcombina-tionHIVtherapyLarderetal.,2007.Inexperimentalsettings,neuralnetworksarealsousedforanalyzingreal-timedataincomputerautomatedstudiesinvolvingrodents.Forexample,inrats,toclassifybehaviorusingFourier-transformed,dig-itizedimagesHeerenandCools,2000;Rousseauetal.,2000ortoidentifysleepstagethroughEEGrecordingsRobertetal.,1997.Human-andneuralnetwork-basedevaluationofratsleepstagewaking,paradoxicalsleep,non-REMusingasingleparietal-occipitalEEGchannelwereinagreementover90%ofepochsexam-inedRobertetal.,1996.Inmice,neuralnetworkshavebeenusedtodistinguishamongvedierentspeciesusingbothfrequency-andtime-domaincharacteristicsofrecordedvocalizationsTianandShang,2006.Surprisingly,athree-layerednetwork57x89x5whichincludedpotentiallyextraneoussignalinputoutperformedafour-layeredtopologyx35x30x5,whichwassuppliedwithasmallersubsetofpowerspectrumdata,withrespecttooverallclassicationapproximately95%versus45%TianandShang,2006;theseeminglyextraneoussignalcharacteristicsprovidedtotherstnetworkevidentlyprovidedimportantdiscriminativedetailsfortheclassier.Indeed,theperformanceofneuralnetwork-basedclassiersisoftenstronglyinuencedbythenumberofattributese.g.,behavioral/cognitivemeasures,recordedsignalssuppliedasinputtothenetworkLeightyetal.,2008.Inadditiontoexperimentalandmedicalapplications,articialneuralnetworkshavedemonstratedutilityforpsychologicalandbehavioralassessmentPriceetal.,2000.Forexample,aneuralnetwork-basedclassierusingcodedresponsesfromparentinterviewsshowedhighdiscriminability92%accuracyduringtrainingand99

PAGE 110

generalizability%correct-assignmentoftestcasesforautismCohenetal.,1993.Nairetal.999usedelevendepression-relatedsymptomstoidentifyclinically-depressedDSM-IIIindividuals.Subsequentcontributionanalysiswhichmeasurestheimpactofeachsymptomonoverallnetwork-produceddiagnosisofthetrainedneuralnetworkidentieddierentpatternsofsymptomimportance"betweende-pressedandnon-depressedindividualsNairetal.,1999,althoughsadness"wasthestrongestindividualpredictor.Thepredictionofpsychiatrictreatmentoutcome,aswell,isapracticalapplicationofneuralnetworks.Serrettietal.07combinedmultiplepredictorsofantidepressantresponseinpatientswithmajordepression,obtainingacorrelationof.46betweenpredictedandobservedresponsei.e.,explain-ing21%ofvariance;thisexampleunderscorestheabilityofneuralnetworkstocombinediverseformsofpredictorvariablese.g.,demographic,individualmedicalandpsychiatrichistory,knowntherapeuticresponseparameters.Inadditiontocross-sectionalstudiesi.e.,samplingconductedatasingletime-point,longitudi-nalanalysesi.e.,samplingatmultipletime-pointsareusefulperformancemetricsfordisorderswithnonlinearpathologicaltrajectories,suchasAlzheimer'sdiseaseTandonetal.,2006.Neuralnetworkshavetheoreticalsignicance,asfunctionalmodelsofbiologicalnervoussystems,forexploringnormalandpathologicalcognitiveprocessesaswellascomplexsystems,machinelearningandarticialintelligence.ANNs,asexamplesofparalleldistributedprocessing,areusedtostudyhumanlearningandproblemsolv-ingRumelhartandMcClelland,1986andarethefoundationofconnectionist"architecturemodelsincognitivesciencee.g.,formodelingtheStroopeect;Co-henetal.,1990.DespitetheANN'scapacityforfault-tolerance,asharpdeclineinoperatingperformanceassociatedwithexternaldisruptionordisconnectionofcon-nectionsbetweencomputingelementsmayoccurwhenupto70%ofconnectionsare100

PAGE 111

compromisedKalampokisetal.,2003;thiscriticalthresholdrepresentstheonsetofpathology"intheANN,reminiscentofincipientneuropathologicaldegenerationofthebrain.Inaddition,computationalmodelsofepisodicmemoryhavebeenbasedonautoassociativeneuralnetworkse.g.,Hopeldnets,1982,andrecognitionmemoryhasbeenstudiedusingbackpropagationinencoderproblem"learningarchitecturesRatcli,1990.3.3.3SupportVectorMachinesSupportvectormachinesSVMsarealearningmethodologyforperformingbi-naryclassicationBurges,1998;CristianiniandShawe-Taylor,2000;Noble,2006.SVMsoriginatedinVladimirVapnik'sVapnikandLerner,1963;VapnikandChervo-nenkis,1964;Boseretal.,1992;CortesandVapnik,1995;Vapnik,1998researchonlearningtheory,postulatingatwo-stageprocessforconstructingabstractpartitionsbetweentwogroupsoftrainingexamples.Intherststage,theinputdatarepre-sentedasnumericalvectorsaremathematicallymappedintoahigher-dimensionalfeaturespace"extendingtheconventionaltwo-orthree-dimensionalspace.Themappingoperationrequiresanappropriatetransformationknownasakernelfunc-tion",whichmustbeselectedinadvancee.g.,polynomial,radialbasisfunction.Inthesecondstage,adecisionsurfacecalledahyperplane"iscalculatedwhichmaximizesseparationbetweenthetwogroups.Attheendoftraining,thetwogroupswilloccupydistinctregionsofthefeaturespace.Likeneuralnetworks,SVMsareveryrobustnoise-tolerant,highlyresistanttoovertting,andcapableofnonlinearclassication.SVMshavebecomeastandardtoolinbioinformaticsresearchforclassifyinggeneexpressionproles.Theexpressionpatternsofcandidategenesareclassiedthroughcomparisonwithpatternsofknownco-regulatedgroupsofgenesGaasterlandand101

PAGE 112

Bekiranov,2000.Golubetal.1999,forexample,distinguishedbetweenindivid-ualswithacutelymphoblasticleukemiaALLandacutemyeloidleukemiaAMLusingSVMstrainedwithAymetrixmicroarraydata17genesfrom38bonemarrowsamplesALL,N=27;AML,N=11.Biologicalsequenceanalysisisanotherimportantbioinformaticsapplicationforsupportvectormachines.Predictionofpro-linecis/transisomerizationSongetal.,2006,forexample,usingadjacentaminoacidsequenceinformationfromaproteindatabase,alongwithmultiplesequencealignmentPSI-BLASTprolesdata;radialbasisfunctionkernelsoutperformedbothpolynomialandlinearkernelfunctionsinthesupportvectormachinestested.SVMsalsoexhibitgreatfacilityforhandlingthevoluminousdatasetsofmeasure-mentssampledduringphysiologicalandbehavioralexperimentsinrodents.Crisleretal.008,forexample,usedasupportvectormachine-basedclassierforsleep-staginginrats.Electrophysiologicaldatawererecordedthreeelectrodes:frontalcor-tex,parietalcortex,andtemporalismuscle,lteredlow-pass,90Hz,anddigitizedtogeneratetime-andfrequency-domaindatasets,whichweresubsequentanalyzedbysupportvectormachineswitharadialbasisfunctionkernel.Scoringresultsbe-tweentheSVM-basedclassierandhumanexpertswerein96%agreementCrisleretal.,2008.Followingtreatmentwithdierentantidepressants,behavioralrecordingstimespentimmobile,swimming,andstrugglingofratsduringtheforcedswim-mingtestwereanalyzedbySVMstodistinguishamongdrugclassesFrohlichetal.,2008.Untreatedcontrolanimalsweredistinguishablefromratsreceivingeithertricyclicantidepressantsimipramine,88%accuracy;desipramine,83%accuracyorselectiveserotoninreuptakeinhibitorsuoxetine,87%accuracy;inaddition,theSVM-basedclassierwasabletodistinguishbetweenanimalsreceivingthetwoclassesofantidepressantsaccuracygreaterthan83%Frohlichetal.,2008.102

PAGE 113

SVMsarepotentiallyusefuladjunctsformedicaldiagnosis.Ubeyli008,forexample,distinguishedamongsixclassesofdermatologicconditionsin358patientsusingasupportvectormachine-basedclassierwitharadialbasisfunctionkernel.Thenetworkcongurationconsistedof34inputsclinicalandhistopathologicalfea-tures,sixoutputsthediagnosticclasses,andninesupportvectors.Theresultingclassierachievednearly100%accuracy,sensitivity,andspecicityUbeyli,2008,outperformingbothmultilayer-perceptronandrecurrentneuralnetworks.Addition-ally,theuseofSVMsindiagnosticimageanalysise.g.,detectingmicrocalcicationclustersindigitalmammograms;El-Naqaetal.,2002underscorestheremarkablepatternrecognitioncapacityofthismethodology.3.4PracticalComparisonsPerformancecomparisonswithinandbetweenstatistical-anddatamining-basedclassiersinvolvesseveralpracticalandtheoreticalconsiderationsDuin,1996.Theparticularapplication,forexample,imposesconstraintsontheselectionofclassi-ers;e.g.,known/unknownunderlyingdistributions,samplesizeeects.Thepriorexperienceofthepractitionerisanimportantfactor,particularlywhenconfrontingdesignheuristics,e.g.,neuralnetworktopologyorsupportvectormachinekernelfunction.Recallingthatthereisnobest"classiermethodologyforallproblemdo-mainse.g.,Duin,1996;WolpertandMacready,1997,comparisonsarenecessarilycontext-specicandperhaps,tosomeextent,user-specic.Patternrecognitionisatypicaldomainforcomparingdiscriminantanalysisandneuralnetworkse.g.,Holmstrometal.,1997.Forinstance,handwrittencharacteridenticationandphonemerecognitionare,respectively,space-andtime-dependentpatternclassicationproblems.Visualimagestimuliprintedcharactersmaybecodedasnormalizedmatricesofpixel-intensityvalues,forexample,andprovidedas103

PAGE 114

inputtotheclassierengine.Similarly,auditorystimulisoundfragmentsmaybeencodedasfrequencyspectra,orassignalprocessingparameterse.g.,calculatedbyFourieranalysis,wavelettransform.ParsonsandJones2000,forexample,usedtemporalandspectralfeaturesextractedfromdigitizedrecordingsofecholocationcallsfrom14sympatricbatspeciesasclassierinput,resultingin79%accuracyofclassicationbydiscriminantfunctionsand87%accuracybyarticialneuralnetwork.Statistical-anddatamining-basedclassiershavebeencomparedinstudiesin-volvingtherapeuticinterventionsinAlzheimer'stransgenicmice,toevaluatebothtransgenicandtreatmenteectsLeightyetal.,2008.Multimetricbehavioraldatafromtwoseparateinvestigations{onestudyexaminingcaeineadministrationinyoungmice,andasecondstudyinvolvingenvironmentally-enrichedhousingcondi-tions{wereanalyzed.Neuralnetwork-basedclassierswereshowntobesuperiortodiscriminantanalysisbothcompleteandoptimized,stepwise-forward,methodsfordistinguishingbetweennontransgenicandAlzheimer'stransgenicanimalsinbothstudies.Thetwoclassiersperformedcomparably,however,withrespecttodistin-guishingtreatmenteectsintransgenicanimalsTgcontrolvs.Tgtreatmentinbothstudies.Inaddition,theclassiersperformedcomparablyfortheinteractionbetweentreatmentandtransgenicityi.e.,discriminabilityamongallgroups.Takentogether,thesendingssuggestthatthetwoclassiersmaybesuitablecomplementsforneurobehavioralinvestigation.Furthermore,neuralnetworksmaybesensitivetosubtlefeaturese.g.,nonlinearityindatasetswhichdiscriminantanalysiscannotdetect.BothstatisticalanddataminingmethodologieshavebeenappliedinAlzheimer'sdiseaseresearch,primarilybypsychologistsandpsychiatristsfordevelopingandre-ningpsychometricscreeninginstruments.Forexample,Grossietal.07usedfourpathologicalfeaturesneurobrillarytangleandneuriticplaquemeasuresin104

PAGE 115

neocortexandhippocampustodistinguishconrmed-ADindividualsN=26andnormal-agedindividualsN=36withdiscriminantanalysisandneuralnetworks.Thelineardiscriminantanalysiscorrectlyidentied92.3%ofindividuals,whiletheneuralnetworkcorrectlyidentiedalltheindividuals.Subsequentrelevanceanal-ysisidentiedtheneocorticalneurobrillarytanglemeasureasthemostimportantpredictorvariableGrossietal.,2007.Inaddition,Frenchetal.997usedthere-sultsofelevenneuropsychologicalteststodistinguishbetweenprobableAlzheimer'sdiseaseandnormal-elderlyindividualswithlineardiscriminantanalysisandneuralnetworks.Theneuralnetworkwasmoreaccuratethanthediscriminantanalysis1.1%versus71.9%overall,andbetterabletodistinguishlevelsofseveritywithintheADindividualsFrenchetal.,1997.Indeed,neuralnetworksarepowerfultech-niquesforcombiningheterogeneouspredictorvariablesinADdiagnosis.Forexam-ple,Garcia-Perezetal.98trainedathree-layeredfeedforward-backpropagationneuralnetworkusing46measuresdemographic,medicalhistory,psychometricas-sessment,electrophysiological,anddiagnosticimagingfrom35individualswithei-therclinically-diagnosedvasculardementiaN=19orADN=16.Subsequenttest-ingusingatrainingsetofdementedindividualsN=23resultedin82.6%accuracyofclassication.Thenetworkcongurationconsistedof46inputlayerelements,29hiddenlayerunits,andasingleoutputunit;thelearningrateandmomentumparameterswerebothsetto0.1,connectionweightswereinitializedto0.3,andthestoppingcriterionerrortolerancewassetto0.0000002Garcia-Perezetal.,1998.105

PAGE 116

3.5What'sPastIsPrologue"Whereofwhat'spastisprologue,whattocomeInyoursandmydischarge.WilliamShakespeare564-1616TheTempest,ActII,Scenei,Lines253-254.TheprecedingchapterswereintendedtoestablishthepracticalfoundationsforconductingbehavioralresearchusingatransgenicmousemodelofAlzheimer'sdis-ease,includingselectionofanappropriateanimalmodelandbehavioral/cognitiveassessmenttaskbattery,aswellastointroducestatisticalanddatamining-basedapproachesforneurobehavioraldataanalysis.Incontrast,thefollowingchaptersillustratespecicapplicationsforexaminingbothtransgenicityandtherapeutictreatmenteectsinnontransgenicandAlzheimer'stransgenicmice,comparingstatisticalanddatamining-basedanalytictechniquesus-ingasemanticinterferencetestingprotocolinhumanAlzheimer'spatients,andforevaluatinganovel,mouse-basedcognitiveassessmentparadigminspiredbythese-manticinterferenceprotocol.Eachapplicationisintroducedwithabriefstatementofbackgroundmotiva-tion,followedbyacompletedescriptionofthematerialsandmethodsutilizede.g.,subjects,treatmentandassessmentprotocols.Next,adetailedpresentationofsta-tisticalanddatamininganalyticresultsisprovided.Thepresentationiscompletedbyadiscussionofkeyndings,emphasizingsignicantcontributionstoAlzheimer'sresearch.106

PAGE 117

CHAPTER4CAFFEINEADMINISTRATIONINNONTRANSGENICMICE4.1IntroductionDoeslife-longconsumptionofcaeineprovideanytherapeuticbenetsinnor-mal,healthyindividuals,withrespecttocognitiveability?Thatis,doescaeineconsumptionsignicantlyprevent,postpone,oramelioratenormalage-associatedcognitiveimpairment?Theeectsoflong-termcaeineconsumptioninnormalhumanshavebeeninves-tigatedthroughretrospectivestudiesoncoeeconsumptione.g.,Chou,1992;Rossoetal.,2008.Approximatelyone-halfoftheAmericanpopulationconsumescoeeonadailybasis,averagingovertwocupsperdayequivalentto200mgofcaeine,or2.4mg/kg/dayforadultsChou,1992.Despitewidespreadanecdotalevidenceofpurportedadversesystemiceectse.g.,anxiety,insomnia,hypertension,coro-naryheartdisease,pancreaticcancer,reproductivesystem-relateddisordersassoci-atedwithchroniccaeineconsumption,thereisnoclearevidenceofalinkbetweenmodestcoeeconsumptionandhumandiseaseChou,1992;Smith,2002.Indeed,self-reportstudiesSmith,2002suggestanassociationbetweenimprovedmentalfunctioningsustainedattention,overallalertnessandregularcoeeconsumptioninnormaladults.Amongelderlywomen,butnotmen,increasedlong-termcoeeconsumptionisassociatedwithbettermentalperformanceMMSE,visualshort-termandlong-termmemory,verbalcategoryuencyRanchoBernardoStudy;Johnson-107

PAGE 118

Kozlowetal.,2002,suggestingacognitiveneuroprotectiveeectofcaeineRossoetal.,2008.Similarly,acommunity-basedFrenchstudyRitchieetal.,2007re-portedrelativelysmallerage-associateddeclinesinverbalretrievalandvisuospatialmemoryabilitywithincreasedlong-termcoeeconsumptioninelderlywomen,com-paredwithage-matchednon-consumers.Inmice,relativelyfewstudieshaveaddressedthelong-termeectsofcaeineadministrationinrodents,despiteanabundanceofstudiesontheacuteeectsofcaeineinmicee.g.,Izquierdoetal.,1979;Angeluccietal.,1999.Forexample,inastudyinvolvingadultratsreceivingcaeineorallythroughtheirdrinkingwater.3mg/mLforfourweeks,caeine-treatedanimalsdisplayedimpairedMorriswatermazeacquisitioni.e.,longerlatencyduringthersttrial,butsimilarlatencyonsub-sequentacquisitiontrials,relativetountreatedratsHunetal.,2007.Additionally,althoughallratsperformedcomparablyonaprobetrialMorriswatermazereten-tionadministeredoneweekafteracquisition,thecaeine-treatedanimalsperformedsignicantlypooreronprobetrialsadministeredtwo-andthree-weeksafteracqui-sitionHunetal.,2007,relativetountreatedrats.Theperformanceimpairmentobservedinthehippocampal-dependentMorriswatermazetaskbycaeine-treatedratswasunderscoredbyimmunohistochemicalassays,whichshoweddecreasedneuro-genesisinthehippocampalsubgranularzoneHunetal.,2007,relativetountreatedanimals.Hence,long-termcaeineadministrationinnormalratsmaybeassociatedwithfunctionaldamagetovulnerableregionsofthehippocampus,expressedbehav-iorallyasdomain-speciccognitiveimpairmente.g.,spatialreferencememory.Bycontrast,astudybyArendashetal.09comparedcognitiveperformancebetweenaged15-16month-oldnontransgenicmicewhichreceivedoralcaeine.3mg/mLfortheprevioustenmonthsandage-matcheduntreatedanimals.AsillustratedinFIgure4.1,cognitivemeasuresobtainedfromacomprehensivebehavioralassess-108

PAGE 119

mentbatteryFigure4.1A,Y-maze;Figure4.1BandC,Morriswatermaze;Figure4.1D,CircularPlatform;Figure4.1E,PlatformRecognition;andFigure4.1F,Ra-dialArmwatermazetasksshowednosignicantdierencesbetweencaeine-treatedanduntreatednontransgenicmiceforanyofthe18behavioralmeasuresevaluatedArendashetal.,2009. Figure4.1.Comparisonsbetweenuntreatedandcaeine-treatednontransgenicmice.AssessingthetreatmenteectofcaeineinnontransgenicanimalsprovidesanimportantreferencebenchmarkforsubsequentevaluationoftherapeuticecacyinAlzheimer'stransgenicmice.Inaddition,thedatasetgeneratedfromsuchinvestiga-tionsprovidesanuniqueopportunityforcomparingconventionalstatisticalanalytic109

PAGE 120

approachese.g.,ANOVAwithnoveldatamining-basedmethodsforidentifyingex-perimentaltreatmenteectsusingamultimetricbehavioralassessmentbattery.Thepurposeofthisstudyistore-examinetheresultsoftheArendashetal.009studyusingdiscriminantanalysisanddatamining-basedclassiers,todeterminewhethergroupwisedierencesincognitiveperformancecanbedetectedthroughadvancedstatisticalanddatamining-basedtechniques.4.2MaterialsandMethodsAtotalofnineteensubjectsnontransgenicmice;C57/SJL/SW/B6hybridback-groundwereincludedinthestudy,randomly-assignedtoeitheroftwogroups:con-trolN=11andcaeine-treatedN=8.Caeinetreatmentprovidedthroughdrink-ingwater,0.3mg/mlbeganatveandone-halfmonthsofage,andcontinueduntilallanimalsofbothgroupsweretestedat15-16monthsofage.Behavioraltestingconsistedofanine-taskbehavioralassessmentbattery,fromwhichnineteenmeasureswereobtainedrefertoTable2.1forcoding:OpenFieldOF,BalanceBeamBB,StringAgilitySA,Y-mazeYM-AE,YM-PA,ElevatedPlusmazeEP-CE,EP-OE,EP-TO,MorriswatermazeWM-Avg,WM-Fin,WM-Ret,CircularPlatformCPE-Avg,CPL-Avg,PlatformRecognitionPR-Avg,PR-Fin,andRadialArmwatermazeRAWM;RML-T4,RML-FT4,RML-T5,RML-FT5.AnimalcareandusewasinaccordancewiththeGuideandUseofLaboratoryAn-imals,NationalResearchCouncil,1996,inaprogramandfacilitiesfullyaccreditedbytheAssociationforAssessmentandAccreditationofLaboratoryAnimalCare,International,underaprotocolapprovedbytheUniversityofSouthFloridaInstitu-tionalAnimalCareandUseCommitteeNo.2729,GaryArendash,Ph.D.,PrincipalInvestigator.110

PAGE 121

GeneralAnalyticProtocolPentiumTM-classmicrocomputinghardwareplatformswereusedtoexecutecur-rentversionsofbothstatisticalanddatamininganalyticalsoftware.SupplementarysoftwareapplicationswereimplementedintheC++programminglanguageStrous-trup,1985fordatabasecross-checking,datapreprocessing,andotherroutinenon-analytictasks.StatisticalanalyseswereperformedusingtheSYSTATTMSystatSoftware,Inc.softwarepackage.Correlationanalysis,factoranalysis,anddiscrim-inantanalysiswereconductedinteractivelyusingdefaultapplicationandsystem-conguredsettings,exceptasindicated.ThePearsonproduct-momentcorrelationcoecientranditsassociatedsignicancevaluepwascalculatedandreportedforeachsignicanti.e.,p<.05pairwisecomparisonbetweenvariables.Bonfer-roni'scorrectionformultiplesimultaneouscomparisonswasapplied,asneeded,andpairwise-deletionwasusedtohandlemissingvalues.Factoranalysiswasperformedusingprincipalcomponentsmethod,withsubsequentVarimaxrotation.Discrimi-nantanalysiswasperformedusingbothcompletedirect-entrymethod,inwhichallavailablepredictorvariablesareusedtogeneratemultivariateclassicationfunc-tions,andstepwise-forwardmethod,whereinamodelisiterativelyconstructedthroughinclusion/deletionofpredictorvariablesonthebasisofindividualvariancecontribution.Thealpha-to-add"parameterforthestepwise-forwardapproachwas0.15.Classierperformancewasevaluatedusingjackkning,andthesuccessrateoverallaccuracy,sensitivity,andspecicitywerecalculatedandreported.111

PAGE 122

Theformulaeforcalculatingaccuracy,sensitivity,andspecicityare:Accuracy=TP+TN TP+TN+FP+FNSensitivity=TP TP+FNSpecificity=TN TN+FPWhere:TPandTNdenotethenumberoftrue-positiveandtrue-negativecases,respectively,andFPandFNrepresentthenumberoffalse-positiveandfalse-negativecases,respectively.Accuracy,sensitivity,andspecicitymeasuresareexpressedaspercentages.Inaddition,Wilks'slambdastatisticWilks,1932andassociatedsig-nicancevaluewerereported.DatamininganalyseswereperformedusingtheJava-basedWekaTMWittenandFrank,2000softwarepackage.Decisiontree-,neuralnetwork-,andsupportvectormachine-basedclassierswereexecutedinteractivelyusingdefaultapplicationandsystem-conguredsettings,exceptasindicated.Deci-siontreeinductionwasperformedusingtheJ48algorithm,aJava-basedimplemen-tationoftheC4.5methodQuinlan,1993.Amultilayerperceptronarchitecturewasusedfortheneuralnetwork-basedclassiers,consistingofaninputlayercorrespond-ingtothenumberofpredictor/attributevariables,anoutputlayercorrespondingtothenumberofdistinctclassicationgroups,andasinglehiddenlayer.Thenumberofcomputingelementsinthehiddenlayerwasvariedbetweenthenumberofinput-layerunitsandthenumberofoutput-layerunits,andthecorrespondingclassierperformancesk-foldcross-validationcomparedtodeterminetheoptimalhidden-layerconguration.Thebackpropagationlearningalgorithmwasusedfortrainingeachneuralnetwork,usingxedparametersforbothlearningrate.9andmomen-tum0.2.ThesupportvectormachineimplementationusedtheSMOmethodandradialbasisfunctionRBFkernelforclassication.Substitutionofanormalized112

PAGE 123

polynomialkernelNPK,insomecases,resultedinimprovedclassierperformance,whereindicated.Multipleexecutionsprogramrunsofbothneuralnetworksandsupportvectormachineswereperformedtoachievestableconvergence,aswellastoavoidspuriousinconsistentresults.Theperformanceofalldatamining-basedclassierswasevaluatedbyk-foldleave-one-outcross-validation,andthesuccessrateaccuracy,sensitivity,andspecicitywerereported,inadditiontotheKappastatisticforcomparisonamongclassiers.4.3ResultsofStatisticalAnalyses4.3.1CorrelationAnalysisSignicantcorrelationsp<.05observedinpairwisecomparisonsofallbe-havioralmeasuresareshowninTable4.1,whereinmarkedcellsincludeboththecorrelationcoecientr-value,topandsignicancep-value,bottom.Signicantintra-taskcorrelationsexistwithintheElevatedPlusmaze,Morriswatermaze,Plat-formRecognition,andRadialArmwatermaze.Signicantinter-taskcorrelationsareobservedbetweensensorimotorandcognitivetasks:OpenFieldandMorriswatermazeOF/WM-Avg,OpenFieldandCircularPlatformerrorsOF/CPE-Avg,andBalanceBeamandRAWMlatencyBB/RML-FT4,RML-FT5,andRML-T5.Indeed,observedintertaskcorrelationsbetweensensorimotorandcognitivemeasuresunderscoretheinterdependenceamongsensory,motor,andcognitivebehavioralpro-cesses.Additionally,signicantinter-taskcorrelationsexistbetweencognitivetasks:Y-mazepercentalternationandRAWMlatencyYM-PA/RML-FT5,RML-T5,CircularPlatformlatencyandMorriswatermazeCPL-Avg/WM-Ret,andPlat-formRecognitionandRAWMlatencyPR-Fin/RML-T4,RML-T5;PR-Avg/RML-T4,RML-T5.Correlationsobservedbetweenmeasuresspanningmultiple113

PAGE 124

taskssuggestsoverlappingorsharedcognitivedomainsacrosstasks,ordependenciestherein.Notsurprisingly,forexample,spatialmemory-biasedtasksMorriswatermaze,PlatformRecognition,andRAWMexhibitextensiveintercorrelationbetweencomponentmeasures.TheabsenceofsignicantcorrelationsbetweenmeasuresfromtheRAWMandMorriswatermazetasksisconsistentwithfactoranalyses,whichindicatedsegregationbetweenthesetasks.Finally,themultifactorialcharacterofcertaintaskswasunderscored:SignicantcorrelationbetweentheElevatedPlusmazetaskandbothRAWMlatencyEP-TO/RML-FT4,RML-T4andMorriswatermazeEP-TO/WM-RetsuggestsananxietycomponenttotheRAWMandMorriswatermazetasks,whileanactivitycomponenttotheCircularPlatformtaskissuggestedbythecorrelationbetweenY-mazeentriesandCircularPlatformerrorsYM-AE/CPE-Avg.4.3.2FactorAnalysisAnunrotatedprincipalcomponentanalysisofthebehavioralmeasuresisshowninTable4.2,withsignicantfactorloadingsabsolutevaluegreaterthan0.540in-dicated.Calculatedeigenvalueresultsi.e.,>1criterionwereinagreementwithobservedscreeplotCattell,1966identicationofvesignicantfactors.Thepri-maryfactoraccountingforover27%ofoverallvariancewasacognitive-biasedstructurecomprisedofthefourRAWMlatencymeasures,aswellastheBalanceBeam,andPlatformRecognitionmeasures,consistentwithcorrelationanalyses.Thesecondfactorwascomprisedofarm-entrycountsbothopen-andclosed-armsfromtheY-maze,aswellastheprobetrialretentionmeasurefromtheMorriswatermaze,reectingtheproportionoftimespentinthepreviouslyplatform-containingquadrantofthepool.Collectively,thisfactormayreectelementsofpersistenceorsearchbehavior.Bothsensorimotoropeneldactivity,Y-mazearmentriesand114

PAGE 125

Table4.1.Correlationsbetweenbehavioralmeasuresinthenontransgeniccaeinestudy BB SA YMAE YMPA EPCE EP1.00 OE .000 EPTO WM.48 Avg .039 WM.67 Fin .002 WM-.49 Ret .032 CPE.50 .71 Avg .028 .001 CPL.73 Avg .000 PRAvg PR.92 Fin .000 RML.54 .63 .58 T4 .017 .004 .009 RML-.48 .46 .83 FT4 .037 .046 .000 RML-.54 -.55 .47 .82 .77 T5 .018 .014 .041 .000 .000 RML-.53 -.49 .75 .86 .90 FT5 .020 .032 .000 .000 .000 OF BB SA YMYMEPEPEPWMWMWMCPECPLPRPRRMLRMLRMLAE PA CE OE TO Avg Fin Ret Avg Avg Avg Fin T4 FT4 T5 115

PAGE 126

Table4.2.Unrotatedfactorcomponentloadingsinthenontransgeniccaeinestudy Factor Measure I II III IV V RML-T5 0.931 RML-FT5 0.873 RML-FT4 0.873 RML-T4 0.863 BB -0.594 PR-Avg 0.564 PR-Fin 0.545 PM-OE -0.847 PM-CE -0.846 WM-Fin 0.552 CPE-Avg 0.690 OF 0.688 YM-AE 0.599 0.652 WM-Avg 0.575 WM-Ret 0.721 CPL-Avg 0.699 YM-PA -0.576 Variance 27.69% 15.14% 14.35% 11.99% 6.87% cognitiveCircularPlatformerroraverage,Morriswatermazeacquisitionmeasuresareincludedinthethirdfactor,whichmayrepresentameshingofspatialrefer-encelearningandexploratorybehavior.Thefourthfactor,whichiscomprisedoftheaverageCircularPlatformlatency,numberofY-mazearmentries,andMorriswatermazeretentionmeasures,mayreectelementsofescapebehavior.Notethattwoofthesetasksinvolveescapescenarios{fromnoxiousstimulitoasaferefugeCircularPlatformorfromthewatertoadryplatformMorriswatermaze{inwhichtheanimalmustlearntoassociateexperimentalcueswithsafety.TheY-mazearmentriesmeasureYM-AEloadsonbothFactorsIIIandIV,underscoringanexploratory/activitycomponenttoothermeasuresinbothofthesefactors.ThefthfactorconsistssolelyoftheY-mazepercentalternations,whichreectssystematic116

PAGE 127

visitationofallarmsintheY-maze.Thisfactormaythereforerepresentgeneralmnemonicfunctionasacognitivedomainseparatefromothersi.e.,workingmem-ory.NotetheabsenceofElevatedPlusmazeopen-armvisitlatencyPM-TOandStringAgilitySAmeasures,indicatingrelativenon-signicanceofthesebehavioralcomponents.4.3.3DiscriminantAnalysisAdirect-entrycompletediscriminantanalysisdidnotyieldsignicantdiscrim-inabilitybetweenthetwogroupsWilkslambda=0.084,p=.7703.However,astepwise-forwardanalysisproducedsignicantdiscriminabilityWilkslambda=0.162,p=.0041with79%classicationaccuracysuccessrate"Jackkningmethod.Theclassierexhibitedasensitivityof75%i.e.,correctindicationoftreatedanimalsandaspecicityof82%i.e.,correctindicationofuntreatedani-mals,withrespecttocaeinetreatmenteect.Indeed,stepwise-forwarddiscrim-inantanalysissurpassedconventionalANOVA-basedtechniquesArendashetal.,2009bydisinguishingbetweenuntreatedandcaeine-treatednontransgenicmice.Theeightvariablesretainedbythestepwise-forwardanalysisincludedbothsensori-motorandcognitivemeasures:OF,BB,YM-PA,WM-Fin,CPE-Avg,CPL-Avg,PR-Avg,andPR-Fin.Thesignicantdiscriminabilitybetweenthetwogroupsachievedthroughacombinationofsensorimotorandcognitivemeasuressuggeststhatcaeinemayexertinuencebeyondthecognitivedomains.117

PAGE 128

4.4ResultsofDataMiningAnalyses4.4.1DecisionTreeAnalysisWM-Ret<=0.117:NT+CAFF3.0WM-Ret>0.117|SA<=3||OF<=98:NT.0||OF>98:NT+CAFF4.0|SA>3:NT9.0/1.0Figure4.2.SampledecisiontreeclassiergeneratedinthenontransgeniccaeinestudyAdecisiontreeclassierconstructedusingtheJ48algorithmvariantofC4.5isdepictedinFigure4.2.Theclassierattemptedtosplitthedatasetusingthreemea-sures,basedontheentropy-reducinginformationgaincapacityofthesepredictors:WM-Ret,SA,andOF.OnlyoneofthemeasuresMorriswatermazeretention,WM-Rethasastrongcognitivebias,whiletheothertworeectsensorimotorintegrativeStringAgilityandexploratoryOpenFieldabilities.Cross-validationk-foldoftheoptimaltreeresultedinonly36%correctclassicationofcasesKappa=-0.34.Thesensitivityofthisclassierwasonly13%andthespecicitywas55%,withre-specttocaeinetreatmenteect.Therefore,althoughthiswastheoptimaldecisiontree-basedclassier,itdidnotsuccessfullydistinguishbetweenthetwogroupsofanimals.4.4.2NeuralNetworkAnalysisAthree-layeredneuralnetwork,trainedusingthebackpropagationlearningal-gorithmrate=0.3,momentum=0.2,wasevaluatedwiththenumberofhidden-layerunitsvariedbetween4and12.Theoptimalperformancewasobtainedfor118

PAGE 129

anetworkcontainingnineteeninput-layerelementsthebehavioralmeasures,vehidden-layerelements,andtwooutput-layerelementsthetwotreatmentgroups.Cross-validationk-foldoftheoptimalnetworkshowed58%correctclassicationofcasesKappa=0.16,with63%sensitivityand55%specicity.Althoughtheseresultsaresuperiortothoseofthedecisiontree-basedclassier,thediscriminabilityobservedisnotsignicant.4.4.3SupportVectorMachineAnalysisAsupportvectormachinearchitectureSMOmethodattemptedtoclassifythenineteencasesintotheirrespectivegroups,resultingin52%correctclassicationKappa=0.08usingk-foldcross-validation,with63%sensitivityand46%speci-city.Theseresultsareinferiortothoseobtainedbytheneuralnetwork-basedclassi-erand,subsequently,representunsuccessfuldiscriminationbetweenthetwogroups.4.5DiscussionThecorrelationanalysisshowedsignicantintercorrelationsbetweenandwithinbehavioraltasks,whichisconsistentwithndingsfromearlierstudiesinvolvingmultimetricbehavioralassessmente.g.,ArendashandKing,2002;Leightyetal.,2004.Thefactoranalysisisconsistentwithndingsfromearlierstudies,inwhichtheprimaryfactorrepresentsastrongcognitive-loadedcomponentofbehavioralperformancee.g.,Galsworthyetal.,2005,largelydrivenbytheRAWMtask,un-derscoringtheprimacyofthistaskforcognitiveassessmente.g.,Leightyetal.,2004;Jensenetal.,2005.Thediscriminantanalysisisalsoconsistentwithearlierstudies,inwhichnon-signicantdiscriminabilityobservedwithdirect-entryanalysisiscontrastedwithsignicantdiscriminabilitywhenthestepwise-forwardapproachisusede.g.,Arendashetal.,2004b;Leightyetal.,2004.Inaddition,thepresenceof119

PAGE 130

bothsensorimotorandcognitivemeasuresintheoptimalmodelsuggestthatcaeineinuencesmultiplebehavioraldomains.Noneofthedatamining-basedclassierssuccessfullydistinguishedbetweenanimalswhichreceivedcaeineandthosewhichdidnot.Table4.3summarizestheperformanceofallclassiersexamined,withre-specttosensitivity,specicity,andoverallsuccessrate,indistinguishingbetweencaeine-treatedanduntreatednontransgenicmice.Table4.3.Classierperformancecomparisoninthenontransgeniccaeinestudy Evaluation DiscriminantAnalysis Decision Neural SupportVector Criterion Complete Step-Fwd Tree Network Machine Accuracy 63% 79% 36% 58% 52% Sensitivity 75% 75% 13% 63% 63% Specicity 55% 82% 55% 55% 46% Althoughthestepwise-forwardvariantofdiscriminantanalysiswasabletodis-tinguishbetweentreatedanduntreatedanimalsusingahighly-selectivesubsetofthebehavioralmeasures,theconsensusamongtheclassiersexaminedisthatthetwogroupsareindistinguishable.Strongagreementamongthedatamining-basedclassi-ersunderscoresthelackofdiscriminabilitybetweenthetwogroupsofmice.Indeed,thecombinationofmeasuresselectedbythestepwise-forwarddiscriminantanalysistoachievediscriminabilitydoesnotexhibitfacevalidity.Uponcarefulinspection,thelistofpredictorvariablesappearsarbitraryandlackscoherencei.e.,itisnotameaningful"samplingofvariables,basedonactualexperiencewithbehavioraltesting.Furthermore,despitethecapacityofneuralnetworkstolearnarbitrarymappingsbetweenfeaturese.g.,behavioralmeasuresandcorrespondinginstancese.g.,treatmentgroupsCybenko,1989,theinabilityofneuralnetwork-basedclas-sierstosuccessfullydistinguishbetweengroupsofanimalsonthebasisofbehavioralmetricsfurtherimpliessimilitude.Theseresultsmirrorthestandardanalysesofvari-120

PAGE 131

ancecomparingcorrespondingperformancemeasuresbetweencaeine-treatedanduntreatednontransgenicmiceintheY-maze,Morriswatermaze,CircularPlatform,PlatformRecognition,andRadialArmwatermazetasksArendashetal.,2009;measureformeasure,theseANOVAsindicatenosignicantdierencesbetweenthetwogroupsofmice.Collectively,thelackofdiscriminabilitybetweencaeine-treatedanduntreatedanimalssuggeststhatchroniccaeineadministrationinnontransgenicmice,begin-ninginearlyadulthood"andcontinuinguntillateadulthood,"doesnotprovidesignicantbenetswithrespecttocognitiveperformance,relativetountreatedani-mals.Oneimplicationforhumansmaybethatlong-term,dailyoralintakeofcaeineinmoderatedosesisunlikelytoprovidecognitivebenetsinnormal,healthyindi-viduals.121

PAGE 132

CHAPTER5CAFFEINEADMINISTRATIONINALZHEIMER'STRANSGENICMICE5.1IntroductionDoeslong-termconsumptionofcaeineprovidecognitivetherapeuticbenetsinindividualsexhibitingAlzheimer'sdiseaseorAlzheimer-likeneuropathology?Theneuroprotectivepotentialofcaeineinhumanshasbeenstudiedthroughcommunity-wideretrospectiveandprospectivestudiesamongcoeeconsumerse.g.,Johnson-Kozlowetal.,2002;Ritchieetal.,2007;vanGelderetal.,2007,whereinincreasedcoeeconsumptionisassociatedwithbothreducedriskforAlzheimer'sdiseaseandsmallerrelativeage-associateddeclinesincognitiveabilities.Indeed,a21yearfollow-upstudyEckelinenetal.,2008suggeststhatmidlifeconsumptionofbetweenthreeandvecupsofcoeesubstantiallyreducestheriskofsubsequentAD.Anotherstudysuggestinganassociationbetweenlong-termcaeineconsump-tionandADMaiaanddeMendonca,2002foundthatADpatientsconsumedlesscaeineduringthe20yearspriortodiagnosis,relativetoage-matchedindividualswithoutAD.However,investigatingthepotentialtherapeuticbenetsofcaeinewithinvulnerablepopulationse.g.,individualspredisposedtoAlzheimer'sdiseaseisamoredicultresearchproblem.Indeed,widespreadmethodologicalissuese.g.,lackofstandardmeasuresforcaeineconsumption,ADdiagnosticcriteriafurthercomplicateeortstoevaluatepossibletherapeuticbenetsofcaeineinADpatientsRossoetal.,2008.122

PAGE 133

Studiesinvolvinglong-termcaeineconsumptioninadultfourmonth-oldAlz-heimer'stransgenicmiceArendashetal.,2006showcognitiveprotectionacrossmultipledomainsspatiallearning,referencememory,workingmemory,andobjectrecognitionfollowingvemonthsoforaltreatment1.5mg/day,relativetoage-matcheduntreatedanimals.Inaddition,long-termcaeineadministrationisassoci-atedwithdecreasedhippocampalbeta-amyloiddepositionandreducedexpressionofbothpresenilinPS1andbeta-secretaseBACEArendashetal.,2006,suggestingreducedproductionofbeta-amyloidincaeine-treatedanimalsasthemechanismofcognitiveprotection.InanotherstudyArendashetal.,2009,theeectsoflong-termcaeineadministrationwereexaminedinaged18-19month-oldAlzheimer'stransgenicAPPswmice.Followingpre-testingintheRAWMtoconrmcognitiveworkingmemoryimpairment,agedtransgenicanimalsconsumedeithercaeine-treated.3mg/mLdrinkingwaterornormal,untreateddrinkingwaterforfourtoveweeks.Age-matchednontransgenicanimals,includedinthestudyforsubsequentgroupwiseperformancecomparisons,consumeduntreateddrinkingwaterthroughoutthestudy.Allanimalsweresubsequentlyevaluatedusingamultimetricbehavioralbattery,andtheresultsanalyzedtodeterminewhetherchroniccaeinedosinginu-encescognitiveperformanceagainstabackgroundofAlzheimer-likepathologyandcognitiveimpairment.Figure5.1illustratescaeinetreatment-associatedreversalofmemoryimpair-mentinagedTgmiceArendashetal.,2009.AgedTgmicedisplaysignicantp<.000005workingmemoryimpairmentintheRAWMtask,relativetoage-matchedNTcontrolanimals,priortocaeinetreatmentFigure5.1-A.AgedTgmicereceivingcaeineTg+CAFFperformedcomparablytoNTmice,withrespecttoRAWMworkingmemorynal-blockT4andT5trials,after4-5weeksofcaeinetreatmentFigure5.1-B,whileTganimalsdiersignicantlyp<.025from123

PAGE 134

theothertwogroups.UntreatedtransgenicmiceTgdidnotexhibitsignicantimprovementbetweenpre-treatmentandduring-treatmentevaluationinRAWMT5performance,althoughbothNTanimalsp<.05;pairedt-testandcaeine-treatedanimalsTg+CAFFp<.005;pairedt-testimprovedmarkedlyFigure5.1-C.ComparableperformanceduringthenaltwodaysofplatformrecognitiontestingstrategyswitchingwasobservedinNTandTg+CAFFmice,whileTgcontrolanimalsshowedimpairmentp<.02;Tgvs.NT.Thepurposeofthisstudywastore-evaluatetheresultsoftheagedAlzheimer'stransgenicmiceexperimentinvolvinglong-termcaeineadministrationArendashetal.,2009,tocomparestandardstatisticalanalysesANOVAwithadvancedstatisti-calanddatamining-basedmethodologiesforidentifyingtransgenic-andtherapeutictreatment-eects.124

PAGE 135

Figure5.1.BehavioralassessmentinNTandTgmicereceivingcaeine125

PAGE 136

5.2MaterialsandMethodsAtotaloftwentysubjectstwelveAPPswAlzheimer'stransgenicmiceandeightage-matchednontransgenicmicewereincludedinthestudy.At18to19monthsofage,allanimalswerepre-treatmenttestedintheRAWMtaskforsixdaystocon-rmworkingmemoryimpairmentintransgenicmice.Afterpre-treatmentRAWMevaluation,thetransgenicanimalswererandomlyassignedtoeitheroftwotreat-mentgroupsTgandTg+CAFF,balancedbyRAWMperformance.FiveofthetransgenicanimalsreceivedcaeineTg+CAFFthroughtheirdrinkingwater.3mg/mL,whiletheothersevenTgcontinuedtoreceivenormaldrinkingwater,asdidthenontransgenicmice,forafour-tove-weektreatmentperiod.Followingthetreatmentperiod,allanimalscompletedathree-taskbehavioralassessmentbattery,fromwhicheightmeasureswereobtainedrefertoTable2.1forcoding:RadialArmwatermazeRAWM;RML-T4,RML-FT4,RML-T5,RML-FT5,PlatformRecogni-tionPR-Avg,PR-Fin,andY-mazeYM-AE,YM-PA.Detaileddescriptionsofthecomputingresourceshardware,software,aswellasparametersettingsforanalyticprograms,areprovidedintheGeneralAnalyticProtocolinSection4.2.AnimalcareandusewasinaccordancewiththeGuideandUseofLaboratoryAn-imals,NationalResearchCouncil,1996,inaprogramandfacilitiesfullyaccreditedbytheAssociationforAssessmentandAccreditationofLaboratoryAnimalCare,International,underaprotocolapprovedbytheUniversityofSouthFloridaInstitu-tionalAnimalCareandUseCommitteeNo.2729,GaryArendash,Ph.D.,PrincipalInvestigator.126

PAGE 137

5.3ResultsofStatisticalAnalyses5.3.1CorrelationAnalysisTable5.1.CorrelationsbetweenbehavioralmeasuresintheAlzheimer'stransgeniccaeinestudy YMPA PRAvg PR.90 Fin .000 RML.65 .52 T4 .002 .020 RML.80 .68 .75 FT4 .000 .001 .000 RML.69 .71 .74 .61 T5 .001 .000 .000 .004 RML.63 .58 .58 .69 .79 FT5 .003 .007 .008 .001 .000 YMYMPRPRRMLRMLRMLAE PA Avg Fin T4 FT4 T5 Signicantcorrelationsp<.05observedinpairwisecomparisonsofallmea-suresforallthreegroupsNT,Tg,Tg+CAFFcollectivelyareshowninTable5.1,whereinmarkedcellsincludeboththecorrelationcoecientr-value,topandsig-nicancep-value,bottom.Signicantintra-taskcorrelationsexistwithinthePlat-formRecognitionandRadialArmwatermazetasks,suggestingextensiveassociationand/orcoordinationamongcognitiveprocesseswithinthesetasks.EectiveoverallcognitiveperformanceinRAWM,forexample,isdependentuponsuccessfulfunc-tionalintegrationbetweenspatiallearningandbothworkingandreferencememorysystems.Widespreadsignicantinter-taskcorrelationswereidentiedbetweenthePlat-formRecognitionandRadialArmwatermazetasks,aswell,suggestingtheinvolve-127

PAGE 138

mentofsimilarunderlyingcognitivedomainsorneurobehavioralsubstratesacrosstasks.Observedinter-taskcorrelations,therefore,suggestthatthePlatformRecog-nitionandRAWMtaskssharemultiplelearning-andmemory-relatedfeaturesincommon.ThecorrelationsbetweencomponentmeasuresacrosstasksinmiceareanalogoustotheobservedcorrelationsbetweenpsychometrictestinginstrumentsusedforhumanADevaluation,resultingfromparallelassessmentofsimilarcogni-tiveabilitiese.g.,workingmemory,spatialrecognition.NeithermeasureobtainedintheY-mazetaskwassignicantlycorrelatedwithanyothermeasures.Thisisnotsurprisingfortheactivity-basedY-mazearmen-triesYM-AEmeasure.TherelativeindependenceofY-mazealternationsYM-PAfromtheothercognitivemeasuresmayreectitsgeneralmnemonicfoundationasseparatefromthecognitivedomainsforRAWMandPlatformRecognitionperfor-manceand/orcontext-sensitivityoflearningandperformance,i.e.,learningwithinaland-basedtaskY-mazeversuswater-basedtasksPlatformRecognition,RAWM.5.3.2FactorAnalysisTable5.2.UnrotatedfactorcomponentloadingsintheAlzheimer'stransgeniccaeinestudy Factor Measure I II PR-Avg 0.902 RML-T5 0.883 PR-Fin 0.854 RML-FT4 0.849 RML-FT5 0.836 RML-T4 0.809 YM-AE -0.868 YM-PA 0.712 Variance 57.11% 17.96% 128

PAGE 139

AnunrotatedprincipalcomponentanalysisoftheeightmeasuresisshowninTable5.2,withsignicantfactorloadingsabsolutevaluegreaterthan0.700in-dicated.Calculatedeigenvalueresultsi.e.,>1criterionwereinagreementwithobservedscreeplotCattell,1966identicationoftwosignicantfactors.Thepri-maryfactor7%ofoverallvariancewascomprisedofallbehavioralmeasuresfromthePlatformRecognitionandRAWMtasks,andrepresentsanoverallcognitivefunc-tioncomponentofbehavior.BothY-mazemeasurestogethercomprisedthesecondfactoraccountingforapproximately18%ofvariance,underscoringtheirindepen-dencefromtheotherbehavioralmeasures,aswasevidentfromcorrelationanalysesaswell.TheY-mazetaskinvolvesbothsensorimotorandcognitiveelements,andthesegregationofthisland-basedtaskfromthetwowater-basedtaskssuggestscontext-sensitivityinbothlearningandperformancelandvs.waterand/orthegeneralmnemonicnatureoftheY-mazetask,asmentionedpreviously.5.3.3DiscriminantAnalysisTheresultsofthediscriminantfunctionanalyses,bothdirect-entrycompleteandstepwise-forwardapproaches,arepresentedinTable5.3forreferencepurposes.Alldiscriminantanalysiscomparisonsinvolvetheeightbehavioralmeasuresfromduring-treatment"testing.NontransgenicNTvs.Transgenic-ControlTgGroupsAdirect-entrycompletediscriminantanalysisdidnotdemonstratesignicantdiscriminabilitybetweenthetwogroups.However,astepwise-forwardanalysisre-turnedsignicantdiscriminabilityWilks'slambda=0.456,p=.0017with87%accuracyJackkningmethod.Thesensitivitywas86%andthespecicitywas88%,withrespecttothetransgenic-controlgroup.OnlytheRML-T5measurewas129

PAGE 140

Table5.3.ClassierperformancecomparisonintheAlzheimer'stransgeniccaeinestudy Support Evaluation DiscriminantAnalysis Decision Neural Vector Groups Criterion Complete Step-Fwd Tree Network Machine Tg Accuracy NS 87% 80% 73% 67% vs Sensitivity NS 86% 100% 71% 43% NT Specicity NS 88% 63% 75% 88% Tg Accuracy NS 75% NS 67% 75% vs Sensitivity NS 80% NS 80% 60% Tg+CAFF Specicity NS 71% NS 57% 86% NT Accuracy NS 100% NS 77% 85% vs Sensitivity NS 100% NS 80% 80% Tg+CAFF Specicity NS 100% NS 75% 88% Allthree Accuracy 60% 80% 45% 70% 65% retainedbythestepwise-forwardmodel,underscoringthesensitivityofthereferencememorymeasuretodistinguishtransgenicity.Transgenic-ControlTgvs.Transgenic+TreatmentTg+CAFFGroupsNonsignicantdiscriminabilitywasobtainedusingdirect-entrydiscriminantanal-ysis.Astepwise-forwardanalysis,however,returnedsignicantdiscriminabilityWilks'slambda=0.233,p=.0065with75%accuracyJackkningmethod.Theobservedsensitivitywas80%andthespecicitywas71%,withrespecttothetransgenic+treatmentgroup.Threevariableswereretainedbythestepwise-forwardanalysis:YM-AE,YM-PA,andRML-FT5.ThissubsetofmeasuresincludesbothsensorimotorthroughtheY-mazemeasuresandcognitiveelements,suggestingamorecomplexbehavioralprolemaydistinguishbetweentreated-anduntreated-transgenicanimals,reectingbothsensorimotorandcognitivecomponents.NontransgenicNTvs.Transgenic+TreatmentTg+CAFFGroupsAlthoughdirect-entrydiscriminantanalysisdidnotresultinsignicantdiscrim-inabilitybetweenthetwogroups,stepwise-forwardanalysisfoundsignicantdis-130

PAGE 141

criminabilityWilks'slambda=0.164,p=.0007with100%accuracyJackkningmethod.Boththesensitivityandspecicityindiceswere100%,withrespecttothetransgenic+treatmentgroup.Thestepwise-forwardanalysisreturnedamodelhavingthreepredictorvariables:YM-PA,RML-FT5,andRML-T5.Thiscognitive-biasedsubsetofmeasuresreectssystematic,coordinatedsearchcapabilityY-mazepercentalternations,aswellasspatialreferencememoryfunction.AllThreeGroupsIncontrasttothepairwisegroupanalyses,adirect-entrydiscriminantanalysisofallthreegroupsshowedsignicantdiscriminabilityWilks'slambda=0.067,p=.0042,butwithonly60%classicatoryaccuracyJackkningmethod.Further-more,astepwise-forwardanalysisalsoreturnedsignicantdiscriminabilityWilkslambda=0.158,p=.0001with80%accuracyJackkningmethod.Thethreepredictorvariablesretainedbytheclassiermodelwere:YM-PA,RML-FT5,andRML-T5.ThisisthesamesubsetofmeasuresretainedbytheNontransgenicvs.Transgenic+Treatment"classiermodel,andsuggeststhatacombinationofsensori-motorandcognitivebehavioralindicesmaybenecessaryforsuccessfuldiscriminationamongnontransgenicandtransgenicanimalseithercaeine-treatedoruntreated.Thecanonicalscoresplotgeneratedfromstepwise-forwarddiscriminantanalysisisdepictedinFigure5.2.Theaxesrepresentthetwofunctionsusedtodistinguishamongthethreegroupsintheanalysis.131

PAGE 142

Figure5.2.Canonicalscoresplotofdiscriminantanalysisforallthreegroups132

PAGE 143

5.4ResultsofDataMiningAnalyses5.4.1DecisionTreeAnalysisNontransgenicNTvs.Transgenic-ControlTgGroupsAdecisiontreeconstructedusingtheJ48algorithmidentiedasinglemeasure,PR-Fin,asthebestdiscriminativeattributeforsplittingthedatasetintodistinctgroups.Subsequentcross-validationk-foldoftheoptimaltreeshowedin80%cor-rectclassicationofcasesKappa=0.61,with100%sensitivityand63%specicity.Transgenic-ControlTgvs.Transgenic+TreatmentTg+CAFFGroupsAlthoughtwoattributemeasures,PR-AvgandYM-PA,wereselectedbythedecisiontree-basedclassierforprovidingoptimaldiscriminabilitybetweenthetwogroups,only50%Kappa=-0.03oftheindividualanimalswerecorrectlyclassi-ed.Hence,theperformanceoftheclassierrepresentschance-levelnonsignicantdiscriminabilitybetweenuntreatedandcaeine-treatedAlzheimer'stransgenicmice.NontransgenicNTvs.Transgenic+TreatmentTg+CAFFGroupsBothPlatformRecognitiontaskmeasuresPR-Fin,PR-Avgwereselectedbythedecisiontreeclassierashavingoptimalcapacityforsplittingthedatasetintotwodistinguishablegroups.However,becauseonly31%ofanimalswerecorrectlyclassiedKappa=-0.52,theperformanceoftheclassierdidnotexceedalevelcomparabletorandomassignmentofindividualstogroups.Thisndingsuggeststhat,whilethePlatformRecognitionmeasuresexhibitmodestbiasbetweenthetwogroups,thecaeine-treatedtransgenicmicedonotsignicantlydierfromthenontransgenicanimals.AllThreeGroupsTheY-mazepercentalternationsYM-PAandbothPlatformRecognitiontaskmeasuresPR-Fin,PR-Avgwereidentiedashavingsucientinformationvalueto133

PAGE 144

splitthedatabaseintothreegroups.Althoughonly45%ofanimalswerecorrectlyassignedtotheirrespectivegroupsKappa=0.16,thismodestlevelofclassierperformanceexceedsthatofrandom-chanceassignment%.5.4.2NeuralNetworkAnalysisNontransgenicNTvs.Transgenic-ControlTgGroupsAthree-layeredneuralnetwork,inwhichthenumberofhidden-layerunitswasvariedbetweenthreeandnine,demonstratedoptimalperformancewhenconguredwitheightinput-layerelementsthebehavioralmeasures,threehidden-layerele-ments,andtwooutput-layerelementsthetwocategorygroups.Subsequentcross-validationk-foldoftheoptimalnetworkshowed73%correctclassicationofcasesKappa=0.46,with71%sensitivityand75%specicity.Transgenic-ControlTgvs.Transgenic+TreatmentTg+CAFFGroupsFourhidden-layerunitswerenecessarytoachieveoptimalclassicationper-formancebetweentreatedanduntreatedtransgenicmice.Theresultingnetworkperformedmodestly,correctlyclassifying67%ofcasesKappa=0.35,with80%sensitivityand57%specicity.Althoughonlytwo-thirdsofindividualswerecor-rectlyidentiedbygroup,theclassiershowedsuperiordetectionofcaeine-treatedAlzheimer'stransgenicmicesensitivity,relativetountreatedTgcontrolsspeci-city.NontransgenicNTvs.Transgenic+TreatmentTg+CAFFGroupsTheneuralnetwork-basedclassierutilizedfourhidden-layercomputingunitstoproduceoptimaldiscriminabilitybetweentreatedanduntreatednontransgenicanimals.Seventy-sevenpercentofindividualanimalswerecorrectlyassignedtotheirrespectivegroupsKappa=0.53.Theclassierdisplayed80%sensitivityand73%specicity.Hence,anidentically-congurednetworkwasbetterabletoaccurately134

PAGE 145

andreliablydistinguish%vs.67%acaeinetreatmenteectintransgenicanimalsagainstanontransgenicstandard,relativetoatransgenicstandard.AllThreeGroupsTheoptimalneuralnetwork-basedclassierfordistinguishingamongthethreegroupscontainedfourcomputingunitsinthehiddenlayer.Thisclassierperformedremarkablywell,demonstrating70%accuracyKappa=0.54,relativetothe33%whichwouldbeattributedtorandomassignment.Theneuralnetwork'sdicultyindistinguishingbetweenuntreatedandcaeine-treatedAlzheimer'stransgenicanimalsTgvs.Tg+CAFF,mentionedpreviously,mayhaveunderminedtheclassier'soverallperformancewhenallthreegroupswereincluded.5.4.3SupportVectorMachineAnalysisNontransgenicNTvs.Transgenic-ControlTgGroupsAsupportvectormachine-basedclassierwastrainedusingthebehavioralmea-surestodistinguishbetweennontransgenicNTanduntreatedtransgenicTgani-mals.Two-thirdsoftheanimalswerecorrectlyclassiedKappa=0.31,with43%sensitivityand88%specicity.Hence,althoughonlymodestdiscriminabilitywasobservedbetweenthetwogroups,theclassierwasparticularlyeectiveincorrectlyidentifyingnontransgenicanimals%specicity,relativetothetransgenicmice.Transgenic-ControlTgvs.Transgenic+TreatmentTg+CAFFGroupsTheclassiercorrectlydistinguishedbetweencaeine-treatedanduntreatedtrans-genicmicein75%ofthecasesKappa=0.47.Thesensitivityofthesupportvectormachinewas60%andthespecicitywas86%,indicatinggreaterdetectionaccuracyforuntreatedmicerelativetotreatedanimals.135

PAGE 146

NontransgenicNTvs.Transgenic+TreatmentTg+CAFFGroupsThesupportvectormachine-basedclassierperformedremarkablywell,correctlyassigning85%ofthemicecomprisingthesetwogroupsKappa=0.68.Theclas-sierdisplayed80%sensitivityand88%specicity.Hence,theclassiershowedap-proximatelyequalfacilityincorrectlyidentifyingnontransgenicandcaeine-treatedtransgenicanimals.AllThreeGroupsWhenasupportvectormachinewastrainedusingbehavioraldatafromallan-imalsofthethreegroups,only65%ofindividualswerecorrectlyassignedtotheirrespectivegroupsKappa=0.45.Overall,thislevelofdiscriminabilityissimilartothatbetweentreated-anduntreated-transgenicanimals.5.5DiscussionConsistentwithndingsfromearlierstudiese.g.,ArendashandKing,2002;Leightyetal.,2004,signicantintercorrelationswerefoundbothbetweenandwithinbehavioraltasks.ThefactoranalysiswasalsoconsistentwithpriorndingsAren-dashandKing,2002;Jensenetal.,2005;Arendashetal.,2006,whereinacognitive-loadedbehavioralcomponentwasreturnedastheprimaryfactor.Additionally,inboththecurrentstudyandinpriorstudiesexamininglong-termcaeineadministra-tioninmiceArendashetal.,2006,measuresfromthePlatformRecognitionandRAWMtasksloadedseparatelyfromtheY-mazetaskmeasures,underscoringthedistinctsensorimotor/cognitivedomainassessmentoftheY-mazeparadigm.Thediscriminantanalysesareconsistentwithearlierstudiese.g.,ArendashandKing,2002;Leightyetal.,2004,aswell,wherenon-signicantdiscriminabilityreturnedthroughdirect-entryanalysisiscontrastedwithsignicantdiscriminabilityachievedbyutilizingthestepwise-forwardapproach.136

PAGE 147

Amongthedataminingclassiers,decisiontreeswereparticularlysensitivetotransgenicityeects,whilesupportvectormachinesweremoresensitivetocaf-feinetreatmenteectsinAlzheimer'stransgenicmice.Thedecisiontreemethodwassuccessfulonlyindistinguishingbetweennontransgenicanduntreatedcontroltransgenicanimals.Therelianceofdecisiontree-basedclassiersuponplatformrecognition-associatedmeasures,however,contrastswithstatisticalclassiers,whichfavoredY-mazeandRAWMmeasures.Thedataminingmethods,whileinferiortothestepwise-forwardvariantofdiscriminantanalysis,werecollectivelysuperiortothedirect-entrycompletediscriminantanalysisclassier.Thesendingssupporttheuseofdataminingtechniquesassupplementsto{ratherthansubstitutesfor{standardstatisticalanalyticmethods.Table5.3summarizestheperformanceofallclassiersexamined,withrespecttosensitivity,specicity,andoverallaccuracy,indistinguishingbetween/amongcaeine-treatedanduntreatedtransgenic,andnon-transgenicmice.Takentogether,theseresultssuggestdierentialsensitivitytotreatmentand/ortransgenicitybypredictorvariables;somevariablesarebetterdiscriminatorsoftransgenic-ortreatment-eectsthanothers.Thisexplainsthesuccessofstepwise-forwarddiscriminantanalysisandfailureofdirect-entrydiscriminantanalysisindistinguishingbetweengroups,intermsoftreatmentaswellastransgenicity.Fur-thermore,thestepwise-forwarddiscriminantanalysessuggestthethreegroupsexam-inedarepairwise-discriminable,albeitutilizingdierentpredictorvariables.ApriorstudyutilizingthesamedatasetasinthepresentstudyArendashetal.,2009em-ployedstandardANOVA-basedcomparisonsbetween/amongnontransgenic,untreated-andcaeine-treatedtransgenicmice,andfoundthatcaeine-treatedtransgenican-imalsdieredsignicantlyfromuntreatedtransgenicsinthePlatformRecognitionescapelatencyandRAWMnalacquisitiontriallatency,T4;retentiontrialla-137

PAGE 148

tency,T5tasksFigure5.1.Inaddition,caeine-treatedtransgenicmicedidnotdiersignicantlyfromnontransgenicsinthesetasksArendashetal.,2009.Al-thoughtheseresultsseemtocontradictthendingsofthecurrentstudyinvolvingdis-criminantanalysisanddatamining-basedmethods,itisimportanttorecognizethatmultiplepredictorvariablesbehavioralmeasureswereutilizedwithineachanalysisofthecurrentstudy,incontrasttothesinglemeasure-by-measurecomparisonsper-formedthroughANOVAinArendashetal.009.Forexample,measuresfromboththeY-mazeandRAWMtasks{together{werenecessarytodistinguishbetweennontransgenicandcaeine-treatedanimalsusingdiscriminantanalysis.Hence,sig-nicantdiscriminability,whenobserved,inthecurrentstudyreectsdistinguishablecollectivepatternsofmeasuresbetweengroups,ratherthandierencesbetweensin-glemeasures.Indeed,theabilityofbothstepwise-forwarddiscriminantanalysisanddatamining-basedclassierstodistinguishbetweengroupsusingmultiplebehavioralmeasuresunderscorestheimportanceoftheseadvancedtechniquesascomplementstoconventionalbehavioralanalyticprotocols,suchasANOVA.Severalimplicationsforhumansmaybeinferredfromthesendings.First,be-causetransgenicmicebothcaeine-treatedanduntreatedweredistinguishablefromnontransgenics,itislikelythatnormalindividualsandAD-individuals,re-gardlessoftherapeuticinuence,willexhibitdistinctivepatternsofcognitiveperfor-mance,howeversubtle,whicharediscernabletoawell-conditionedclassierusingasucientlyrichdatasetofmeasures.Thissupportstheuseofcomprehensivebehav-ioralandcognitiveassessmentforaccurate,reliablediagnosisofprobableAlzheimer'sdisease.Second,inlightofndingsbyArendashetal.09thatlong-termoralcaeineconsumptionconferscognitive-protectivebenetsinagedAlzheimer'strans-genicmicebasedonspecictaskperformancemeasures,itisreasonabletoconcludethatcertaincognitivedomainse.g.,working/referencememoryarehighlyrespon-138

PAGE 149

sivetothetherapeuticeectsofcaeine.Finally,thatcertaintasksaremorelikelytodetectthesecognitivebenets,whenpresent,thanareothertasks.TheRadialArmwatermaze,forinstance,hasdemonstratedsuperiorityforidentifyingcognitiveimpairment,aswellasforevaluatingtheecacyoftherapeuticinterventions,inAlzheimer'stransgenicmicee.g.,Arendashetal.,2004;Jensenetal.,2005;Ethelletal.,2006;Cracchioloetal.,2007.Similarly,theMMSEFolsteinetal.,1975psychometricinventoryisoneofthemosteectiveinstrumentsforbothdiagnosticscreeningandtherapeuticevaluationinAlzheimer'spatients.Assessmentmethod-ologiesforAlzheimer'sdiseasecontinuetoimprove,however,andrecently-developedparadigmse.g.,semanticinterferencetask;Loewensteinetal.,2004emphasizespe-ciccognitivedomainse.g.,verbalmemoryandoperationse.g.,proactiveandretroactiveinterferencewhichexhibitAlzheimer-specicimpairment,asexploredinthenextchapter.139

PAGE 150

CHAPTER6INTERFERENCETESTINGINHUMANS:ACOMPARISONOFSTATISTICALANDDATAMININGMETHODS6.1IntroductionMemorydysfunction{particularly,impaireddelayedrecallability{isanearlyindicatorofprobableAlzheimer'sdiseasee.g.,Welshetal.,1991.CompromisedstorageandconsolidationofnewlearninginADpatientsmayberelatedtoincreasedsusceptibilitytostimulusintrusionsoccurringbetweenthelearningperiodandsub-sequentrecalli.e.,interferencewhich,inturn,isassociatedwithimpairedhip-pocampalfunctione.g.,HasselmoandWyble,1997.Anovelcognitiveassessmentprotocol,developedbyDr.DavidLoewensteinandcolleaguesattheUniversityofMiamiFloridaSchoolofMedicine,hasdemonstratedeectivenessasapsychome-tricscreeninginstrumentfordistinguishingamongclinically-diagnosedDALCOG,MMSEmildprobableAD,mildcognitivelyimpairedMCI,andnormalelderlyindi-vidualsLoewensteinetal.,2003;Loewensteinetal.,2004.ThetechniquerepresentsanextensionoftheFuldobjectrecognitiontaskusingtwosetsofsemantically-relatedstimuliforevaluatingbothproactiveandretroactiveinterferenceeects.Theinterferencetestingprotocolconsistsoffourtasks.Thersttask,three-trialrecall,isamodiedversionoftheFuldobjectmemoryexaminationFuld,1981;Loewensteinetal.,2001,inwhichthesubjectispresentedwithtenfamiliarobjectsBagAandaskedtorecalltheobjectsfollowingabriefdistractiontask,repeatedthreetimes.Inthesecondtask,proactiveinterference,thesubjectispresented140

PAGE 151

withtennovelobjectsBagBandaskedtorecallthem,todeterminewhetherpreviouslearningBagAobjectsintrudesuponpresentlearningBagBobjects.Thethirdtask,short-delayrecall,whereinthesubjectisaskedtorecalltheoriginalsetoftenitemsBagA,providesameasureofretroactiveinterferencedicultyrecallingpreviouslearningduetointrusionbypresentlearning.Finally,long-delayrecallisevaluatedbyaskingthesubjecttorecalltheoriginalsetoftenitemsBagAaftera20-minutedelay.Thestimuluspresentationandresponseformswereselectedbothtofacilitateadministrationofthetestinginstrumentaswellastocloselymirrorfamiliar,commonlyexecutedcognitivetaskse.g.,objectrecognitionbytouch,namingbyhumansubjects.Inaddition,theobjectscomprisingthetwosetsofstimuliwereselectedtobesemanticallyrelatedLoewensteinetal.,2003,toincreasethelikelihoodofintrusioninterferenceerrors.BothAlzheimer'sdiseasepatientsandindividualswithmildcognitiveimpairmentwereshowntobevulnerabletoproactiveinterference,relativetonormalelderlysubjectsLoewensteinetal.,2004.Thepurposeofthisstudywastoexploreadvancedstatisticalanddataminingmethodologiesfordistinguishingamongaged-normal,mildAlzheimer's,andmildcognitivelyimpairedindividualsusingmeasuresfromthefourtasksdescribedaboveincludedinthesamedatasetasintheoriginalLoewensteinetal.004study.TheresultsprovidedintheoriginalreportLoewensteinetal.,2004,basedonone-wayanalysisofvarianceANOVAandlogisticregressionstatistics,werecomparedwithdiscriminantanalysesanddatamining-basedanalysesoftheoriginaldataset.141

PAGE 152

6.2MaterialsandMethodsAtotalof132casesrepresentingthreeclassesofage-matchedsubjectswerein-cludedinthedataset:MildAlzheimer'sdiseaseAD,N=26,mildcognitiveimpair-mentMCI,N=53,andnormalelderlyCON,N=53.GroupmembershipwasassignedbyclinicalexaminationusingstandardizedcriteriaNINCDS-ADRDA.Inadditiontobehavioraltesting,allindividualscompletedtheMMSEexamination.TheMMSEscorewasincludedinthecorrelationanalysistoillustrateconvergentvalidity,butonlybehavioralmeasureswereusedforsubsequentanalyses.Thebehav-ioralevaluationyieldedthefollowingmeasures,inorder:Three-trialrecallModiedOME,FuldscoreBag-Aobjects,proactiveinterferenceBag-BImmediateRecallscore,retroactiveinterferenceBag-AShortDelayscore,anddelayed-recallBag-A20-minDelayscore.Averbaluencytaskwasusedasadistractorbetweensuccessivetrialsofthethree-trialrecalltask,aswellasimmediatelyprecedingtheproactiveinterferencetask.ThesedatawereprovidedcourtesyofDr.DavidLoewen-stein;completedetailsofthetestingprocedureareprovidedinLoewensteinetal.004.6.3ResultsofStatisticalAnalyses6.3.1StandardAnalysisFigure6.1depictssignicantgroupwisecontrastsincomponentbehavioralmea-suresofthesemanticinterferenceprotocol,withrespecttomemoryperformancei.e.,averagenumberofcorrectresponses.UsinganalysisofvarianceANOVA,F-test,allthreepairsofgroupsi.e.,ADvs.MCI,ADvs.CON,andMCIvs.CONexhibitedsignicantdierencesincognitiveperformance.Boththethree-trialrecallandproactiveinterferencemeasuresLoewenstein'sFRET3andSITBmeasures,re-142

PAGE 153

spectivelydieredsignicantlybetweenthesethreepairsofgroupsallp=.000.Inaddition,boththeretroactiveinterferenceanddelayed-recallmeasuresLoewen-stein'sSITABandSITDREmeasures,respectivelydieredsignicantlybetweentheCONgroupandeithertheADorMCIgroupallp=.000;thesetwomeasuresalsodieredbetweentheADandMCIgroupsbothp=.020.Furthermore,signicantgroupwisedierenceswereobservedintheaverageMMSEscores[MeanSEM:AD3.120.53,MCI.040.26,andCON8.560.19],whereinallpairwisedierencesbetweengroupswerestatisticallysignicantallp=.000.Hence,foreachofthecognitiveperformancemeasures,allpairsofgroupsdiersignicantlyp<.05and,collectively,portrayacontinuumofcognitiveimpairmentfromrelatively-unimpairedperformanceCONtosubstantialimpairmentAD.NormalelderlyindividualsCON,forinstance,demonstratedsu-periorperformance,relativetotheothertwogroups.Bycontrast,mildlycognitively-impairedMCIindividualsexhibitedintermediateperformancebetweentheothertwogroups.6.3.2CorrelationAnalysisTable6.1.CorrelationsamongMMSEscoresandbehavioralmeasuresinthehumansemanticinterferenceprotocol Three-Trial .54 Recall .000 Proactive .46 .75 Interference .000 .000 Retroactive .41 .68 .51 Interference .000 .000 .000 Delayed .41 .78 .60 .75 Recall .000 .000 .000 .000 MMSE Three-Trial Proactive Retroactive Recall Interference Interference 143

PAGE 154

Figure6.1.Groupwisecontrastsforallhumaninterferenceprotocolmeasures144

PAGE 155

Signicantp<.05pairwisecorrelationsamongallclinicalmeasuresbothtask-basedandMMSEforallgroupsofsubjectsareshowninTable6.1.BoththePearsonproduct-momentcorrelationcoecientr-valueanditscorrespondingsignicancep-valueareindicatedatthetopandbottom,respectively.Positivecorrelationbe-tweencognitiveperformanceinthethree-trialrecalltaskandboththeretroactiveinterferenceanddelayed-recallmeasuressuggestsperformanceconsistencybetweentheoriginallearningcomponenti.e.,therstsetofverbalstimuliandsubsequentretroactiveinterference.Inaddition,positivecorrelationbetweentheproactivein-terferencemeasureandtheotherthreeverbaltasksthree-trialrecall,retroactiveinterference,anddelayed-recallsuggestscomparablelearningperformanceacrossproblemconditionsi.e.,dierentsetsofverbalstimuli.Finally,signicantpositivecorrelationbetweentheMMSEandalltask-basedbehavioralmeasuresunderscoresthediagnosticutilityoftheMMSEpsychometricinstrumentforverballearningandmemoryevaluation.6.3.3DiscriminantAnalysisTable6.2comparestheperformanceoftheclassiersexaminedinthepresentstudy,includingthedirect-entryandstepwise-forwarddiscriminantanalyses.Alzheimer'sADvs.MildCognitiveImpairmentMCIGroupsDirect-entrycompletediscriminantanalysisdemonstratedsignicantdiscrim-inabilitybetweenindividualswithprobableAlzheimer'sdiseaseandthosewithmildcognitiveimpairmentWilks'slambda=0.730,p=.0001,with70%ofsubjectscorrectlyassignedtotheirrespectivegroups,asshowninTable6.2.Thesensitivityoftheclassierwas65%andthespecicitywas72%,withrespecttoAlzheimer'sidentication.Stepwise-forwardanalysisalsoreturnedsignicantdiscriminabilitybetweenthetwogroupsWilks'slambda=0.754,p=.0000with76%accuracy,145

PAGE 156

Table6.2.Classierperformancecomparisoninthehumansemanticinterferenceprotocol Support Evaluation DiscriminantAnalysis Decision Neural Vector Groups Criterion Complete Step-Fwd Tree Network Machine AD Accuracy 70% 76% 73% 68% 73% vs Sensitivity 65% 69% NS NS NS MCI Specicity 72% 79% 93% 77% 91% AD Accuracy 94% 94% 91% 90% 95% vs Sensitivity 85% 85% 85% 85% 89% CON Specicity 98% 98% 94% 93% 98% MCI Accuracy 90% 90% 83% 88% 91% vs Sensitivity 87% 87% 83% 87% 93% CON Specicity 92% 92% 83% 89% 89% Allthree Accuracy 71% 81% 73% 73% 76% 69%sensitivityand79%specicity.Thestepwise-forwardmodelretainedonlythethree-trialrecallscorei.e.,Fuldtaskscoremeasure.Hence,thethree-trialre-calltaskcomponentprovidessucientpredictiveinformationtodistinguishbetweenmildcognitiveimpairmentandprobableAlzheimer'sdisease.Indeed,theadditionalbehavioraltaskmeasuresmay,infact,underminediscriminabilitybetweenthesetwogroups.Alzheimer'sADvs.Normal-AgedCONGroupsSuccessful,andidentical,discriminabilitybetweennormalagedandprobableAlzheimer'sindividualswasachievedusingdirect-entryandstepwise-forwarddis-criminantanalysesWilks'slambda=0.245,p=.0000.Bothclassiervariantsexhibited94%accuracy,with85%sensitivityand98%specicity,withrespecttoAlzheimer'sidentication,asreportedinTable6.2.Thethree-trialrecall,proactiveinterference,andretroactiveinterferencetaskscoreswereretainedbythestepwise-forwardmodel.Equivalentperformancebetweenthedirect-entryandstepwise-146

PAGE 157

forwardanalysessuggeststhedelayed-recallmeasuredoesnotcontributetooveralldiscriminabilitybetweennormal-agedandprobableAlzheimer'sindividuals.MildCognitiveImpairmentMCIvs.Normal-AgedCONGroupsDirect-entryandstepwise-forwarddiscriminantanalysesperformedidentically,distinguishingbetweennormal-agedandmildlycognitivelyimpairedindividualsWilks'slambda=0.438,p=.0000with90%accuracy,asindicatedinTable6.2.Theob-servedsensitivitywas87%andthespecicitywas92%,withrespecttotheMCIgroupofsubjects.Thestepwise-forwardmodelretainedthethree-trialrecall,proac-tiveinterference,andretroactiveinterferencetaskscores.Theseresultsonceagainsuggestthedelayed-recallmeasuredoesnotcontributetodiscriminabilitybetweennormal-agedandMCIindividuals.AllThreeGroupsIndividualsofallthreegroupsnormal-aged,probableAlzheimer'sdisease,andmildlycognitivelyimpairedweresuccessfullydistinguishedusingdirect-entrydis-criminantanalysisWilks'slambda=0.337,p=.0000.Seventy-onepercentofthesubjectswerecorrectlyidentiedbygroup.Thenormal-agedindividualswerethemostreliablydistinguishedbytheclassier.Astepwise-forwardanalysisalsore-turnedsignicantdiscriminabilityamongthethreegroupsWilks'slambda=0.220,p=.0000with81%accuracy.Allfourbehavioralmeasureswereretainedbythestepwise-forwardanalysisaspredictorvariablesforgroupmembership.Theseresultssuggestthatallofthecognitivemeasuresmaycontributesignicant,nonredundantdiscriminativeinformationfordetermininggroupmembership.147

PAGE 158

6.4ResultsofDataMiningAnalyses6.4.1DecisionTreeAnalysisAlzheimer'sADvs.MildCognitiveImpairmentMCIGroupsTwobehavioralmeasuresproactiveinterferenceanddelayed-recallwereselectedbythedecisiontreeforsplittingthedatasetKappa=0.3102.Theresultingclassi-ercorrectlyidentied73%ofindividuals,withnonsignicant%sensitivityand93%specicity,asshowninTable6.2.MostofthesubjectswereclassiedasMCI,regardlessofactualgroup.Alzheimer'sADvs.Normal-AgedCONGroupsThedatasetattributesidentiedbythedecisiontreewerethethree-trialrecallanddelayed-recallmeasures.Theclassieraccuratelyidentied91%ofindividualsKappa=0.7974.Thesensitivitywas85%andspecicitywas94%,withrespecttoprobableAlzheimer'sdisease.Hence,thenormal-agedindividualsweresomewhatmorelikelytobecorrectlyidentied,relativetoAlzheimer'ssubjects.MildCognitiveImpairmentMCIvs.Normal-AgedCONGroupsAdecisiontreeattemptedtosplitthedatasetusingthethree-trialrecallandproactiveinterferencemeasures.Theresultingclassiercorrectlyidentied83%Kappa=0.6604ofindividuals,with83%sensitivityand83%specicity,asin-dicatedinTable6.2.AllThreeGroupsThedecisiontree-basedclassierselectedthreemeasuresthree-trialrecall,proac-tiveinterference,anddelayed-recalltodistinguishamongallthreegroupsofindi-viduals.Theoptimaltreeaccuratelyidentied73%oftheindividualsKappa=0.5678.Theretroactiveinterferencemeasurewasnotselectedbyanyofthedeci-siontree-basedclassiers.148

PAGE 159

6.4.2NeuralNetworkAnalysisAlzheimer'sADvs.MildCognitiveImpairmentMCIGroupsAneuralnetwork-basedclassierwastrainedtodistinguishbetweenindividualsfromtheprobableAlzheimer'sdiseaseandmildlycognitivelyimpairedgroupsusingallfourbehavioralmeasuresfromtheinterferenceparadigm.Optimalperformancewasachievedwithathree-layerednetworkcontainingfourcomputingelementswithintheinputlayerelementsthebehavioralmeasures,threehidden-layerelements,andtwooutput-layerelementsthetwocategorygroups.AsindicatedinTable6.2,theclassiercorrectlyidentied68%ofcasesKappa=0.2763,withnonsignicant0%sensitivityand77%specicity,withrespecttoAlzheimer'sdisease.True-ADindividualswereassignedequallytoeithertheADorMCIgroup.Alzheimer'sADvs.Normal-AgedCONGroupsAsshowninTable6.2,aneuralnetwork-basedclassiersuccessfullydistinguishedbetweennormalagedindividualsandthosewithprobableAlzheimer'sdiseaseKappa=0.7707,with90%accuracy,85%sensitivityand93%specicity,withrespecttoAlzheimer'sdisease.Althoughthisclassierexhibitedremarkablyaccurateperfor-mance,15%ofprobableAlzheimer'sdiseaseindividualsweremisclassiedasbeingnormalaged.MildCognitiveImpairmentMCIvs.Normal-AgedCONGroupsDiscriminabilitybetweennormalagedandmildlycognitivelyimpairedindividu-alswasdemonstratedbyaneuralnetwork-basedclassierKappa=0.7547,whichaccuratelyassigned88%ofindividualstotheirrespectivegroups.Theclassiershowed87%sensitivityand89%specicity,withrespecttomildcognitiveimpair-ment.Thus,excellentdiscriminabilitywasachievedbetweenMCIandnormal-agedindividualsthroughneuralnetworkanalysis.149

PAGE 160

AllThreeGroupsAllthreegroupsnormalaged,mildlycognitivelyimpaired,probableAlzheimer'sdiseaseofindividualsweresuccessfullydistinguishedKappa=0.5657,asreportedinTable6.2.Seventy-threepercentofindividualswerecorrectlyassignedtotheirrespectivegroupsusingtheclassier.6.4.3SupportVectorMachineAnalysisAlzheimer'sADvs.MildCognitiveImpairmentMCIGroupsAsupportvectormachinearchitectureattemptedtodistinguishbetweenmildlycognitivelyimpairedindividualsandthosewithprobableAlzheimer'sdiseaseusingthebehavioralmeasuresfromtheinterferenceparadigm.AsshowninTable6.2,theclassiercorrectlyidentied73%oftheindividualsbygroupKappa=0.3253,withnonsignicant39%sensitivityand91%specicity.Hence,althoughalmostalloftheMCIindividualswerecorrectlymatchedtotheirgroup,mostoftheADindividualswereincorrectlyidentiedasbelongingtotheMCIgroup.Alzheimer'sADvs.Normal-AgedCONGroupsNinety-vepercentofnormalagedandprobableAlzheimer'sindividualswerecorrectlyassignedtotheirrespectivegroupsusingasupportvectormachine-basedclassierKappa=0.8830.Theobservedsensitivitywas89%andthespecicitywas98%,withrespecttoAlzheimer'sdisease.Thisclassierperformedremarkablywelloverall,inthatonly11%ofprobableAlzheimer'sindividualswereincorrectlyidentiedasbeingnormalaged.MildCognitiveImpairmentMCIvs.Normal-AgedCONGroupsNormalagedandmildlycognitivelyimpairedindividualsweresuccessfullydistin-guishedKappa=0.8113usingasupportvectormachinearchitecture.Asreported150

PAGE 161

inTable6.2,theclassieraccuratelyidentied91%oftheindividualsfromthesetwogroups,with93%sensitivityand89%specicity,withrespecttoMCI.AllThreeGroupsArespectable76%ofindividualsfromallthreegroupswerecorrectlyassignedtotheirrespectivegroupKappa=0.6110usingclassiersbasedonsupportvectormachinearchitectures.6.5DiscussionTheoriginalreportLoewensteinetal.,2004presentedANOVA-basedstatisti-calresultsofcomparisonsbetweengroupsofalltaskmeasuresintheinterferenceprotocol.Thethreegroupswerefoundtodiersignicantlyallp<.001,withrespecttocognitiveperformance,inallfourtasksthree-trialrecall,proactiveinter-ference,retroactiveinterference,anddelayed-recall.Inaddition,pairwisecompar-isonsbetweengroupswereperformedbylogisticregressionforalltasksexceptthethree-trialrecallcomponentoftheprotocol.ThemildcognitivelyimpairedMCIandnormal-agedCONgroupsshowedsignicantdierencesallp<.0001intheproactiveinterference,retroactiveinterference,anddelayed-recallmeasuresLoewen-steinetal.,2004.Proactiveinterferencewasthebestdiscriminatorbetweenthesetwogroups81.3%accuracy,whiledelayed-recallmeasurewastheleast-eectivepredictorvariable.5%.BetweenthemildADpatientsADandnormal-agedCONindividuals,theproactiveinterferencep<.0001,retroactiveinterferencep<.001,anddelayed-recallp<.0001measuresdieredsignicantlybetweenthetwogroupsLoewensteinetal.,2004.However,betweentheADandCONgroups,delayed-recallprovidedthebestdiscriminability.9%andretroactiveinterfer-enceshowedthepoorestdiscriminability3.3%,ofthefourtasks.Additionally,higheraccuracy,sensitivity,andspecicitywereobservedwhenmultiplemeasures151

PAGE 162

werecombinedasasingleindexofcognitiveperformancee.g.,TotalRecognitionMemoryScoreLoewensteinetal.,2004.Inthisstudy,thecorrelationanalysisshowedsignicantallp=.000posi-tivepairwisecorrelationamongallfourbehavioralmeasures,aswellasbetweenthestandardpsychometricscreeninginstrumentMMSEandeachbehavioralmeasure.Becauseonlyfourbehavioralmeasureswereexamined,andtheseexhibitedexten-siveintercorrelation,nofactoranalyseswereperformed.Thediscriminantanalyseswereconsistentwithearlierstudiese.g.,ArendashandKing,2002;Leightyetal.,2004,inwhichthestepwise-forwardvariantexhibitscomparableor,sometimes,enhancedaccuracy,relativetothestandarddirect-entrycompleteanalysis.In-deed,althoughalldiscriminantanalysis-basedclassierssuccessfullydistinguishedbetweenpairsofgroups,bycognitiveimpairment,thestepwise-forwardapproachwassuperiorfordistinguishingbetweenADandMCIgroups,aswellasamongallthreegroups,asshowninTable6.2.Overall,allthreedecisiontree-basedclassi-erssuccessfullydistinguishedbetween/amongthegroups,butthesupportvectormachine'sperformancewasthemostsimilartothestepwise-forwarddiscriminantanalysisresult.Supportvectormachine-basedclassiersdemonstratedsuperiordiscriminabilitybetweenthemildcognitivelyimpairedMCIandnormal-agedCONindividuals,asdepictedinTable6.2.Thestepwise-forwarddiscriminantanalysisretainedthethree-trialrecall,proactiveinterference,andretroactiveinterferencemeasuresfordistinguishingbetweenthegroups.ThisndingisconsistentwiththatofLoewen-steinetal.04,whereinthesesamethreemeasuresprovidedsuperiordiscrim-inability,relativetodelayed-recall,betweenthesetwogroups.Thedecisiontree,aswell,retainedthethree-trialrecallandproactiveinterferencemeasures,underscor-ingtherelativeimportanceofthesecognitivemetricsfordetectingsubtlefeaturesof152

PAGE 163

MCI.Indeed,Loewensteinetal.04identiedproactiveinterferenceasprovidingoptimaldiscriminability1.3%betweenthetwogroups,bylogisticregression.Thesupportvectormachine-basedclassieralsodisplayedthebestdiscriminabil-itybetweenprobableADandnormal-agedindividuals,asshowninTable6.2.Asnotedearlier,equivalentclassierperformancedespitetheabsenceofthedelayed-recallmeasurefromthestepwise-forwardanalysissuggeststhismeasuredoesnotcontributesignicantlytogroupwisediscriminability.Thiscontrastswiththelogisticregression-basedresultsLoewensteinetal.,2004whichemphasizethedelayed-recallmeasurefordistinguishingbetweenthesetwogroups.Indeed,theoverallaccuracyachievedbydiscriminantanalysis,asshowninTable6.2,exceedsthatoflogisticregressionbetweentheADandCONgroups.Thedecisiontree-basedclassierde-terminedthatthethree-trialrecallanddelayed-recallmeasurestogetherprovidethemostinformationbiasfordistinguishingbetweenthesetwogroups.Theob-servedaccuracyofthisclassier1%,fromTable6.2approximatesthelogisticregression-basednding9.9%fordelayed-recallalone,asreportedinLoewensteinetal.2004.Theneuralnetwork-basedclassierperformedcomparablytothelo-gisticregressionmethod,withrespecttodiscriminabilitybetweenprobableADandnormal-agedCONindividuals.AsindicatedinTable6.2,thestepwise-forwarddiscriminantanalysisdisplayedthebestdiscriminabilitybetweenprobableADandmildcognitiveimpairment,uti-lizingonlythethree-trialrecallmeasure.AlthoughLoewensteinetal.004didnotreportrelativediscriminabilityusingtheFuldthree-trialrecallmeasure,theyemphasizetheclinicalsignicanceofthistaskforthediagnosisofdementia.In-terestingly,thedecisiontreeretainedboththeproactiveinterferenceanddelayed-recallmeasures,andclassiedalmostallindividualsasMCI,regardlessoftruegroupmembership.Indeed,thedicultyofallthreedataminingtechniquestoaccurately153

PAGE 164

distinguishbetweenthesegroupsalldatamining-basedclassiersdisplayednon-signicantsensitivity,eectivelyassigningtrue-ADindividualsrandomlytoeithergroupunderscoresthechallengeofreliablebehavior-baseddierentialdiagnosisofAD.Theadverseeectsofdisproportionatesamplesizesontrainingtheclassierscannotbediscounted,however.AlthoughtheMCIandnormal-agedgroupshadequalsizesN=53,therewereonlyone-halfasmanyADindividualsN=26in-cludedinthestudy.EitherdoublingthenumberofADsubjectsorhalvingthenumbersofMCIandnormal-agedsubjectsbyMonteCarloselection,forexamplemightcorrectthesamplesize-biasproblem.Discriminantanalysesretainedallbehavioralmeasuresfordistinguishingamongthethreegroups,anddisplayed81%overallaccuracy,asreportedinTable6.2.Asnotedearlier,theretentionofallfourbehavioralmeasuresbythestepwise-forwardanalysissuggeststhateachmeasuremaycontributeuniquediscriminativeinforma-tion.Indeed,theLoewensteinetal.2004ndingsindicatethatmultiplemeasuresofcognitiveabilityprovidesuperiordiscriminability,relativetoindividualmeasures.Thedecisiontree-basedclassierretainedallmeasuresexceptretroactiveinterfer-ence,butonlyidentied73%oftheindividualsbygroup,asshowninTable6.2.Takentogether,theseresultssuggesttheinterference-basedcognitiveassessmentprotocolintroducedinLoewensteinetal.04ishighlyeectiveindistinguishingbetweenandamongmild-Alzheimer's,mildcognitivelyimpaired,andaged-normalindividuals.Inaddition,signicantpositivecorrelationbetweenthebehavioralmea-suresandtheMMSEstandardclinicalassessmentinstrumentsupportsconvergentvalidityfordiagnosis.AsportrayedinTable6.2,thereisstrongconsensusamongtheclassierswithrespecttodiscriminabilitybetweengroups,e.g.,ADindividualsaremoreeasilydistinguishedfromaged-normalthanareMCIindividuals.Thecom-parisonsandcontrastsbetweenconventionalstatisticalapproachesANOVA,logistic154

PAGE 165

regression,asreportedinLoewensteinetal.04,andtheadvancedstatisticaldiscriminantanalysesanddatamining-basedapproachesutilizedherefurthersup-porttheuseofthelattertechniquesinconjunctionwiththeformerforcomprehensiveneurobehavioralresearch.Futurestudiesshouldaddressrenementsandextensionsofbothstatisticalanddataminingmethodstoimprovediagnosticutility.155

PAGE 166

CHAPTER7THEINTERFERENCETASK:ANOVELASSESSMENTPARADIGM7.1IntroductionAsdiscussedinthepreviouschapter,thesemanticinterferencetaskLoewensteinetal.,2004isarecently-developedclinicaldiagnosticprotocolfordistinguishingamongnormal-aged,mildcognitively-impaired,andprobableAlzheimer'sindividu-alsusingaverbal-reportmemorytesthavingproactiveandretroactiveinterferencecomponents,aswellasameasureofdelayed-recallperformance.Thedemonstratedeectivenessofthisprotocolfordetectingand,indeed,discriminatingmildformsofcognitiveimpairmentinhumanssuggestedthatasimilarensembleofrodent-basedbehavioraltasksandmeasuresmightexhibitcomparableutility.Thenotionofadapt-inghuman-basedtasksforcognitiveassessmentinrodentsisnotwithoutprecedent.Indeed,themostcommonsingleevaluationinstrumentisthemazee.g.,HebbandWilliams,1946,ascaledmodeloftheclassicpath-searchpuzzle.Bothwater-andland-basedmazes,featuringtwo-andthree-dimensionalcongurations,havebeendevelopedforstudyinglearningandmemoryprocessese.g.,Spearetal.,1990invariedcontexts.Similarly,theserialreactiontimetestforproceduralimplicitmemoryfunctionalevaluationNissenandBullemer,1987hasalsobeenadaptedformicee.g.,ChristieandHersch,2004;Choetal.,2007.Finally,episodic-likememoryfunctionhasbeenexaminedinmiceusingexplorationandobjectrecogni-tiontasksDereetal.,2005tosimulatethewhere,what,andwhen"paradigms156

PAGE 167

commonlyusedinhumans.MirroringtheepisodicmemoryimpairmentdisplayedbyAlzheimer'spatients,Alzheimer'stransgenicmiceperformpoorlyinthesimulatedepisodic-likememorytaskSavonenkoetal.,2005.Thepurposeofthisstudywastoimplementandevaluateanovelbehavioraltestingparadigmformice,adaptedfromthehuman-basedinstrumentpresentedinLoewensteinetal.04,innontransgenicandAlzheimer'stransgenicanimalshav-ingeitherthewildtypeorGRK5-knockoutgenotype.Themouse-analogueoftheinstrument,describedinMaterialsandMethodsbelow,substitutesspatialmemory-dependentelementsfortheverbalsemanticmemory-dependentcomponentsoftheoriginalprotocol,whilepreservingtheexibilityofthelatterforevaluatingproac-tiveandretroactiveinterferenceeectsonlearning.Additionally,incontrasttotheLoewensteinetal.004whichutilizedonlyerror-scores,themouse-basedadap-tationprovidesbotherror-scoreandresponse-latencymeasuresfromeachtask.AsshowninpriorstudieswithbothnontransgenicandAlzheimer'stransgenicanimalse.g.,Arendashetal.,2001;Leightyetal.,2004;Arendashetal,2006,theRadialArmwatermazeRAWMprovidesanexcellentmeansforevaluatingspatialshort-termworkingmemoryfunctioninmice.Consequently,theRAWMrepresentsareasonablestartingdesignfromwhichtodevelopthemouse-basedinterferencetest-ingparadigm.ThetwoalternativecongurationsoftheRAWMapparatusPoolAandPoolB,featuringuniquetargetplacementandperipheralvisualcues,forexam-ple,areintendedtoparalleltheoriginalprotocol'suseofsemantically-relatedobjectsBagAandBagB.Similarly,theintertrialY-mazeexposureperiodisintendedasabriefdistractortask,analogoustotheinterveningverbaluencynamingtaskusedbyLoewensteinetal.2004.G-proteincoupledreceptorkinase-5GRK5isresponsibleforselectivedesensi-tizationofG-proteincoupledreceptorsofthemuscarinicacetylcholinergicsystemin157

PAGE 168

micereviewedin:Suoetal.,2007.Studiesinvitroshowsub-thresholdconcen-trationsofsolublebetaamyloidreducefunctionalGRK5levelsand,moreover,thatthesereductionsprecedetheonsetofcognitiveimpairmentinADtransgenicmiceSuoetal.,2004.Aged19month-oldGRK5-knockoutmiceexhibitwidespreadhippocampalpathologyswollenaxonalclusters,aswellasmuscariniccholinocep-tivesystemdysfunctionreducedmAChRM1,M2,andM4Suoetal.,2007;theseanimalsalsodisplayselectiveimpairmentinworkingmemory,butnotspatiallearningorreferencememory.TheGRK5-knockoutmousehasbeenproposedasamodelorganismforhumanAD,asacomplementtoexistingAPP-transgenicmice,toevaluatebothbehavioralcognitiveandpathologice.g.,cholinergicresponsestotreatmentprotocols.Inthisstudy,themouse-basedinterferenceparadigmwasusedtocomparecon-ventionalstatisticalANOVAmethods,asreportedinSuoetal.07,withad-vancedstatisticaldiscriminantfunctionanddatamining-basedanalytictechniquesforevaluatingcognitiveimpairmentinmicebothnontransgenicandAlzheimer'stransgenicwith/withoutGRK5manipulationi.e.,wildtypeorknockout.Inaddi-tion,error-scoreandresponse-latencydatawereexaminedbothseparatelyandto-gethertodeterminethemosteectivegroupofbehavioralmeasuresfordistinguishingbetween/amongtreatmentgroupsofmice.Hence,theinteractionbetweentwoge-neticmanipulationsnontransgenicvs.Alzheimer'stransgenic,GRK5-wildtypevs.knockout,asreectedincognitiveperformanceintheinterferenceparadigm,wasinvestigated.158

PAGE 169

7.2MaterialsandMethodsTheanimalsusedforthisanalysisrepresentedanextensionoftheinitialbe-havioralstudyreportedin:Suoetal.,2007,involvingone-year-oldAlzheimer'stransgenicAPPswTg2576miceandnontransgeniclittermates,withanadditionalmanipulationoftheGRK5genewildtypevs.knockout.Targeteddeletionofex-ons7and8oftheGRK5genewasusedtogenerateGRK5-KOknockout"mice.Hence,therewerefourgroupsofanimalsexamined:Nontransgenic-controlNT,N=7,nontransgenic-knockoutNT-KO,N=8,transgenic-controlTg,N=12,andtransgenic-knockoutTg-KO,N=10.Allanimalscompletedthecomprehensivebe-havioraltaskbatteryasusedinpriorstudies,e.g.,Jensenetal.,2005;Cracchioloetal.,2007,priortotheMouseInterferenceParadigm.Allbehavioralmeasuresfromthecomprehensivetaskbatterywereanalyzedus-ingbothstatisticalANOVA,correlationanalysis,factoranalysis,anddiscriminantanalysisanddatamining-basedmethodsdecisiontrees,neuralnetworks,andsup-portvectormachines.Similarly,thethreedatasetsofinterferenceparadigm-derivedbehavioralmeasuresobtainedinthisstudyweresubsequentlyanalyzedusingthesameanalyticsuite.Acompletedescriptionofthecomputingresourcese.g.,hard-wareplatform,softwarepackagesusedintheanalyses,includingparametersettingsfortheprograms,isprovidedintheGeneralAnalyticProtocolofSection4.2.AnimalcareandusewasinaccordancewiththeGuideandUseofLaboratoryAn-imals,NationalResearchCouncil,1996,inaprogramandfacilitiesfullyaccreditedbytheAssociationforAssessmentandAccreditationofLaboratoryAnimalCare,International,underaprotocolapprovedbytheUniversityofSouthFloridaInstitu-tionalAnimalCareandUseCommitteeNo.2951,GaryArendash,Ph.D.,PrincipalInvestigator.159

PAGE 170

MouseInterferenceParadigmApparatus.TheapparatusconsistsoftwocircularpoolsA"andB";100cmdiameterx25cmdeep,eachcontainingasix-armedcmlongx19cmwide;radiallydistributedarounda40cmdiametercentralarenastainlesssteelinsert,andarelocatable,transparentsubmergedplatformcmdiameter;e.g.,invertedclearglassjar.Thepoolsarelledwithwarmwatermaintainedat24-27degCtoalevel1.5cmabovethetopofthesubmergedplatform,positionedintheappropriategoal-arm.Foreachpool,thearmsarelabeledinaxed,counter-clockwiseorder,beginningwiththearmlocateddirectlyinfrontoftheexperimenter#1.Adistinctvisualcueuniquesizeandshapeisplacedattheendofarms#2through#6,alongtheexteriorcircumferenceofthepools.Inaddition,aY-mazeapparatusthree-armed;21cmlongx4cmwidex40cmhighwallsislocatednearby.Protocol.Behavioraltestingtypicallyconsistsof4-6consecutivedailysessions,witheachtestday'ssessioncomprisedof:onePlatformAOrientation"trial,threePlatformARecall"trials,onePlatformBOrientation"trial,onePlatformBRecall"trial,onePlatformAShortDelay"trial,andonePlatformALongDe-lay"trial.Inaddition,eachpoolisassignedaspecicstart-armandgoal-armtobeusedthroughouttheday'ssession.Eachpool'sdailystart-armandgoal-armassign-mentsforuptosixtestingsessionsare:A2,6B,6;A,3B,3;A,5B2,1;A,1B,4;A,4B,6;andA,2B,3.Atthebeginningofeachsession,asubmergedplatformispositionedneartheendofeachpool'sgoal-arm.Forthesin-gle,unscoredPlatformAOrientation"trial,themouseisintroducedatthecenterofthestart-armofPoolA,facingthecentralarena,andallowed60sectoexplorethepoolundisturbed.Iftheanimalhasnotlocatedthesubmergedplatform,itisgentlyguidedto{andallowedtoremainatop{theplatformfor30sec.ThemouseisthenplacedintoonearmoftheY-mazeandallowedtoexplorefor60sec.Next,160

PAGE 171

oneachofthreerecalltrials:themouseisintroducedintothestart-armofPoolA,midwayalongthelengthandfacingthecentralarena,andallowedupto60sectolocateandmountthesubmergedplatform.Ifthemouseentersanon-goalarm,itisgentlywithdrawntothestart-armandanerrorisrecorded.Ifthemousefailstolocatetheplatformafter60sec,itisgentlyguidedthereandallowedtoremainfor30sec.Aftereachrecalltrial,themouseisplacedintotheY-mazeandallowedtoexplorefor60sec.Inaddition,foreachrecalltrial,thenumberoferrorstravellingmorethan20cmintoanynon-goalarmandthelatencytimetoreachtheplat-form;scoredassec,"ifunreachedarerecordedasPlatformARecall-Errors"andPlatformARecall-Latency,"respectively.Themouseisthenintroducedatthecenterofthestart-armofPoolB,facingthecentralarena,andallowed60sectoexplorethepoolundisturbedforthesingle,unscoredPlatformBOrientation"trial.Iftheanimalhasnotlocatedthesubmergedplatform,itisgentlyguidedto{andallowedtoremainatop{theplatformfor30sec.ThemouseisthenplacedintoonearmoftheY-mazeandallowedtoexplorefor60sec.Next,forasingletrial,themouseisintroducedintothestart-armofPoolB,midwayalongthelengthandfacingthecentralarena,andallowedupto60sectolocateandmountthesub-mergedplatform.Eachtimethemouseentersanon-goalarm,itiswithdrawntothestart-arm.Ifthemousefailstolocatetheplatformafter60sec,itisgentlyguidedthereandallowedtoremainfor30sec.ThemouseisthenplacedintotheY-mazefor60sec.ThenumberoferrorsandlatencyarerecordedasPlatformBRecall-Errors"andPlatformBRecall-Latency,"respectively.TheanimalisthenplacedintoPoolAforasingletrial,asdescribedabove,andthePlatformAShortDelay"numberoferrorsandlatencyarerecorded.Themouseistransferredtoitshomecagefor20min,afterwhichitisplacedintoPoolAforasingletrial,asdescribedabove,andthePlatformALongDelay"numberoferrorsandlatencyarerecorded.161

PAGE 172

Dataset.Correspondingbehavioralmeasuresobtainedfromthenaltwodailytestingsessionsareaveragedasablock"forsubsequentanalysis.Hence,eightbe-havioralmeasuresaredetermined:Three-trialrecallmeanerror-scoreandresponse-latency,fromPlatformARecall;proactiveinterferencemeanerror-scoreandresponse-latency,fromPlatformBRecall;retroactiveinterferencemeanerror-scoreandresponse-latency,fromPlatformAShortDelay;and,delayed-recallmeanerror-scoreandresponse-latency,fromPlatformALongDelay.Finally,thesemeasuresaregroupedintothreedatasets:Error-scoresonly,response-latenciesonly,andbotherror-scoresandreponse-latencies.7.3ResultsofStatisticalAnalyses:ComprehensiveTaskBattery7.3.1StandardBehavioralAnalysisTable7.1portraysthestandardANOVAanalysesF-test,=.05ofbehavioralmeasuresfromthecomprehensivetaskbatteryforallanimalsinthestudy.Signif-icantdierencesbetweengroupsareprintedinboldanddenotedbysuperscriptedcharacters:Thenontransgenic-controlNTvs.Alzheimer'stransgenic-controlTggroupcontrastisrepresentedbyanasterisk;theAlzheimer'stransgenic-controlvs.transgenicwithGRK5-knockoutgenotypeTg-KOcontrastisindicatedwithadaggery;and,thecontrastbetweennontransgenic-controlandnontrans-genicwithGRK5-knockoutgenotypeNT-KOisrepresentedbyadouble-daggerz.SignicantdierenceswerefoundbetweentheNTandTggroupsinY-mazepercentalternationsYM-PA,butnotYM-AE,elevatedplusmazeopen-armentriesEP-OE;butnotEP-CEorEP-TO,andradialarmwatermazeretentionmemoryoverallerrorsandlatenciesinbothT4andT5;errorsinnalblockT5.TheprimarycognitiveimpairmentinTgmicerelativetoNTanimals,therefore,was162

PAGE 173

reectedintheRAWMerrormeasures,althoughaTgeectonY-mazealternationsgeneralmnemonicfunctionwasalsoevident.TheTgmicedieredsignicantlyfromtheTg-KOanimalsinbothopeneldactivityOFandelevatedplusmazeopen-armentries.Similarly,NTanimalsandNT-KOmicealsodieredsignicantlywithrespecttoopeneldactivity.Hence,nocognitiveeectsoftheGRK5-knockoutgenotypewereobservedineithernontransgenicorAlzheimer'stransgenicmice.NosignicantdierenceswerefoundbetweentheNTandTg-KOgroupsforanyofthebehavioralmeasures.163

PAGE 174

Table7.1.GroupwisecontrastsforallbehavioraltaskmeasuresintheGRK5study Behavioral Groupmeanstandarderror Measure NT Tg Tg-KO NT-KO Sensorimotor-based OF 136.113.5 156.36.7 142.0y11.0 87.6z10.5 BB 53.07.0 39.47.2 35.88.6 39.79.0 SA 4.30.7 4.60.4 3.50.8 3.90.7 EP-CE 12.41.4 15.22.8 11.72.4 11.31.8 EP-OE 0.70.5 30.314.6 10.5y5.4 1.40.5 YM-AE 33.32.7 36.92.9 36.73.1 25.12.4 Anxiety-based EP-TO 0.70.5 1.10.4 1.00.5 3.81.9 Cognitive-based YM-PA 67.05.1 51.02.6 50.33.3 62.63.0 WM-Fin 20.16.7 21.13.3 22.64.4 24.54.7 WM-Avg 28.04.8 28.63.1 31.23.5 33.73.7 CPE-Fin 12.24.1 24.35.2 29.88.1 17.64.9 CPE-Avg 18.13.5 23.54.2 30.16.2 17.53.0 CPL-Fin 122.437.3 166.135.0 172.029.9 166.529.5 CPL-Avg 181.329.2 188.424.2 200.423.7 219.622.0 PR-Fin 13.46.0 12.34.3 7.71.2 18.77.4 PR-Avg 19.75.8 19.32.6 16.02.0 27.66.4 RME-FT4 1.00.6 2.10.3 1.50.5 0.40.3 RME-FT5 1.60.6 3.20.6 2.50.6 0.80.4 RME-T4 1.30.4 2.60.2 2.00.3 1.30.1 RME-T5 1.30.4 2.70.3 2.10.4 1.20.2 RML-FT4 17.97.6 32.03.8 21.66.6 10.43.7 RML-FT5 20.87.1 36.05.6 31.36.9 12.43.3 RML-T4 21.94.1 33.92.6 29.34.9 21.52.0 RML-T5 21.34.4 32.92.8 30.64.9 19.12.1 164

PAGE 175

7.3.2CorrelationAnalysisTable7.2portrayssignicantpairwisecorrelationsp<.05observedbetweenbehavioralmeasuresfromthecomprehensivetaskbattery.Markedcellsincludeboththecorrelationcoecientr-value,topandsignicancep-value,bottom.Widespreadintra-taskcorrelationsexistwithintheRAWMtaskand,toalesserextent,withinthePlatformRecognitionandMorriswatermazetasks.Inter-taskcorrelationswerefoundbetweensensorimotorandcognitivetasks:OpenFieldac-tivityandRAWMallmeasuresexceptRML-T4,CircularPlatformerrors,Y-mazearmentries,andElevatedPlusmazeclosed-armentries;and,StringAgilityandEl-evatedPlusmazeopen-armlatency.Thesepatternsunderscoretheinterdependenceamongsensory,motor,andcognitiveprocesses,aswellastheroleofanxiety-relatedcomponentsonbehavior.Inaddition,signicantinter-taskcorrelationswerefoundbetweencognitivetasks:Y-mazeandallRAWMmeasuresexceptbetweenYM-AEandRME-T5;MorriswatermazeandPlatformRecognition;3MorriswatermazeandRAWMlatenciesand,toalesserextent,errors;and,CircularPlatformlatencyandRAWMerrorsand,toalesserextent,latencies,aswell.Theseobservedcorrela-tionsbetweenbehavioralmeasureswhichspanmultipletasksissuggestiveofshared,overlapping,orinterdependentcognitivedomainsacrosstasks.Forexample,exten-siveintercorrelationbetweencomponentmeasuresoftheMorriswatermaze,Plat-formRecognition,andRAWMtasksreectsthesharedspatialmemorydependencyofthesebehavioralparadigms.Finally,patternsinpairwisecorrelationsunderscorethemultifactorialcharacterofcertaintasks.SignicantcorrelationbetweentheElevatedPlusmazeclosed-armentriesandbothCircularPlatformmeasures,aswellasbetweenElevatedPlusmaze165

PAGE 176

open-armentriesandRAWMerrors,suggestsananxiety/exploratorycomponentwithinboththeCircularPlatformandRAWMtasks.7.3.3FactorAnalysisAvarimax-rotatedprincipalcomponentanalysisofallbehavioralmeasuresfromthecomprehensivetaskbatteryisshowninTable7.3,withsignicantfactorload-ingsabsolutevaluegreaterthan0.500indicated.Theeigenvalueresultsi.e.,>1criterionwereconsistentwithscreeplotCattell,1966identicationofsevensig-nicantfactors.Theprimaryfactoraccountingforabout28%ofoverallvariancewasacognitive-biasedstructurecomprisedofallRAWMmeasures,aswellastheY-mazepercentalternations,consistentwithcorrelationanalysis.ThesecondfactorwascomprisedofPlatformRecognitionandMorriswatermazemeasures,represent-inganobjectrecognition/identication-relatedcognitivedomain,distinctfromtheworkingmemory-associatedprimaryfactor.FactorIIIincludesacombinationofsen-sorimotorBalanceBeamlatencyandstimulus-avoidancemeasuresaverageCircu-larPlatformlatencyandElevatedPlusmazeclosed-armentries,andmayreectanescape/avoidancebehavioralcomponent.Similarly,thefourthfactorcontainsmea-suresassociatedwithescapeoravoidancebehavioraverageCircularPlatformerrorsandElevatedPlusmazeclosed-armentries,aswellasanexploratory/activitymea-sureOpenFieldactivity,whichtogethermayreectescape/avoidancebehaviorsdistinctfromFactorIII.Thefthfactorincludesonlyasinglebehavioralmeasure,ElevatedPlusmazeopen-armresidencytime,whichistheonlyElevatedPlusmazemeasurelinkedtoanxiety;thus,FactorVrepresentsananxietyfactor.FactorVIalsoincludesonlyasinglemeasure,ElevatedPlusmazeopen-armentries,whichmayberelatedtoexploratorybehavior,distinctfromFactorV.Thesixthfactoriscomprisedofasinglesensorimotormeasure,StringAgility,whichrepresentsoverallphysical166

PAGE 177

Table7.2.CorrelationsbetweenbehavioralmeasuresofthecomprehensivetaskbatteryintheGRK5study BB SA YM.44 AE .007 YMPA EP.35 CE .036 EP-.33 OE .050 EP-.34 TO .040 WMAvg WM.90 Fin .000 WMRet CPE.42 .35 Avg .012 .036 CPL-.39 Avg .019 PR.60 .62 Avg .000 .000 PR.65 .68 -.33 .89 Fin .000 .000 .050 .000 RML.45 -.43 .50 .63 .36 T4 .005 .008 .001 .000 .032 RML.40 .41 -.34 .47 .61 -.40 .75 FT4 .014 .011 .038 .003 .000 .015 .000 RML.37 .35 -.45 .57 .60 .73 .76 T5 .023 .036 .005 .000 .000 .000 .000 RML.42 .49 -.48 .39 .41 -.35 .64 .70 .84 FT5 .010 .002 .002 .017 .012 .036 .000 .000 .000 RME.37 .50 -.49 .35 .42 -.34 .84 .65 .59 .59 T4 .022 .002 .002 .035 .010 .046 .000 .000 .000 .000 RME.43 .37 -.41 .37 .52 -.41 .66 .96 .68 .67 .65 FT4 .008 .023 .013 .023 .001 .012 .000 .000 .000 .000 .000 RME.39 -.48 .40 .42 .53 -.39 .62 .68 .90 .74 .60 .65 T5 .018 .003 .015 .010 .001 .019 .000 .000 .000 .000 .000 .000 RME.42 .49 -.51 .44 -.37 .53 .62 .74 .90 .57 .60 .77 FT5 .010 .002 .001 .006 .027 .000 .000 .000 .000 .000 .000 .000 OF BB SA YMYMEPEPEPWMWMWMCPECPLPRPRRMLRMLRMLRMLRMERMERMEAE PA CE OE TO Avg Fin Ret Avg Avg Avg Fin T4 FT4 T5 FT5 T4 FT4 T5 167

PAGE 178

Table7.3.Varimax-rotatedfactoranalysisofcomprehensivetaskbatterymeasuresintheGRK5study Factor Measure I II III IV V VI VII RML-T5 0.877 RME-T5 0.862 RML-FT5 0.826 RME-FT4 0.820 RML-FT4 0.815 RME-FT5 0.779 YM-PA -0.665 RML-T4 0.663 RME-T4 0.629 PR-Fin 0.911 PR-Avg 0.882 WM-Fin 0.812 WM-Avg 0.810 CPL-Avg -0.743 BB 0.719 EP-CE 0.511 0.535 CPE-Avg 0.772 OF 0.768 EP-TO 0.883 EP-OE 0.757 SA -0.746 Variance 27.99% 15.61% 8.45% 9.11% 6.26% 5.74% 6.66% 168

PAGE 179

strength,coordination,andgripcapacity.Interestingly,thesegregationofmeasuresfromtheElevatedPlusmazetaskunderscoresthemultifactorialcharacterofthisbehavioraltask,whichincludesactivityandescape/avoidanceanxiety-related.Table7.4.Classierperformancecomparisonusingcomprehensivetaskbatterymea-suresintheGRK5study Support Evaluation DiscriminantAnalysis Decision Neural Vector Groups Criterion Complete Step-Fwd Tree Network Machine NT Accuracy NS 84% 74% 89% 74% vs. Sensitivity NS 92% 75% 92% 83% Tg Specicity NS 71% 71% 86% 57% NT Accuracy NS 87% NS 67% 53% vs. Sensitivity NS 100% NS 63% NS NT-KO Specicity NS 71% NS 71% 57% Tg Accuracy NS 81% 71% 52% 62% vs. Sensitivity NS 67% 67% NS NS Tg-KO Specicity NS 92% 75% 58% 75% NT Accuracy NS 81% NS 56% NS vs. Sensitivity NS 89% NS NS NS Tg-KO Specicity NS 71% NS 71% NS Allfour Accuracy 47% 56% 36% 39% 28% 7.3.4DiscriminantAnalysisTable7.4comparestheperformanceofclassiersutilizingbehavioralmeasuresfromthecomprehensivetaskbattery,fordistinguishingbetweenpairsofgroups,aswellasamongthefourgroupsofanimals.Thethirdandfourthcolumnsofthetabledisplaytheresultsofdiscriminantanalysis-basedclassiersusingthedirect-entrycompleteandstepwise-forwardmethods,respectively.Evaluationcriteriaforclassierperformanceareprovidedonlywhentheobservedlevelstrictlyexceedsexpectedvaluesassociatedwithrandom-chanceassignmentofindividualstogroupsi.e.,50%fortwo-groupdiscriminability,25%forfour-groupdiscriminability.169

PAGE 180

Nontransgenic-ControlNTvs.Transgenic-ControlTgGroupsAlthoughthedirect-entrycompletediscriminantanalysisfailedtodistinguishbetweentheNTandTggroups,thestepwise-forwardapproachreturnedsignicantdiscriminabilityWilks'slambda=0.191,p=.0003betweenthegroups.Asindi-catedinTable7.4,theclassierexhibited84%accuracy,with92%sensitivityand71%specicity,withrespecttotransgenicity.Fivebehavioralmeasureswereselectedonthebasisofvariancecontribution:SA,YM-PA,CPE-Avg,RME-T4,andRME-T5.Theemphasisoncognitivemeasures,particularlytheRAWMworkingmemoryT4andT5components,underscoresthecognitiveimpairmentofTgmice,relativetoNTcontrols.Nontransgenic-ControlNTvs.Nontransgenic-KnockoutNT-KOGroupsDirect-entrydiscriminantanalysiswasunabletodistinguishbetweentheNTandNT-KOgroups,howeverthevariance-optimizingstepwise-forwardapproachper-formedremarkablywellWilks'slambda=0.317,p=.0044.Threepredictorvari-ableswereselectedforthemodel{OF,EP-TO,andRME-FT5{whichrepresentacross-sectionofsensorimotor,anxiety-related,andcognitivemeasures.Eighty-sevenpercentofindividualswerecorrectlyidentiedbygroup,includingalloftheNT-KOanimalsand71%oftheNTcontrols.ThissuggeststhattheabsenceofGRK5expres-sioninanontransgenicanimalproducesasubtle,albeitdiscernable,characteristicbehavioralphenotype.RelativetoNTmice,NT-KOanimalsshoweddecreasedlo-comotoractivityOF,lessanxietygreaterEP-TOlatency,andsuperiorworkingmemoryfewernal-blockRAWMtrialT5errors.Transgenic-ControlTgvs.Transgenic-KnockoutTg-KOGroupsAsshowninTable7.4,thestandarddirect-entryapproachdidnotsuccessfullydistinguishbetweenthetwogroups,althoughthestepwise-forwardanalysisreturnedsignicantdiscriminabilityWilks'slambda=0.610,p=.0116usingonlytwobe-170

PAGE 181

havioralmeasures,BBandRME-T4.Theclassierdisplayed81%overallaccuracy,with67%sensitivityand92%specicity,withrespecttotheTg-KOgroup.Hence,theoverallperformancewaslargelydrivenbytheclassier'sidenticationoftheTgcontrolgroup.Interestingly,asensorimotormeasurebalancebeamlatencyandarobustcognitivemeasureRAWMoverallT4errors,together,werefoundtoprovideoptimaldiscriminability,althoughasshowninTable7.1thesetwomeasures,indi-vidually,donotdiersignicantlybetweenthegroups.TheTg-KOmicedisplayedbetterbalance/coordinationhigherBBandbettercognitivefunctionfeweroverallRAWMtrialT4errors,comparedwithAlzheimer'stransgenic-controlanimals.Nontransgenic-ControlNTvs.Transgenic-KnockoutTg-KOGroupsThedirect-entrydiscriminantanalysisdidnotdistinguishbetweenthetwogroups,althoughthestepwise-forwardapproachwassuccessfulWilks'slambda=0.199,p=.0028.Theclassierdisplayed81%accuracy,with89%sensitivityand71%speci-city,withrespecttotheTg-KOgroup.FivebehavioralmeasuresBB,EP-TO,YM-PA,PR-Fin,andRML-FT4wereincludedinthemodel.Asdescribedearlier,withreferencetoTable7.1,thesetwogroupsdonotdiersignicantlywithrespecttoanyofthesemeasures.Moreover,theclassierislikelyexhibitingsensitivitytoacompositebehavioralphenotype,whichemergesonlythroughsamplingmul-tiplesensorimotor,anxiety,and/orcognitivefeaturesutilizingthecomprehensivetaskbattery.Relativetonontransgenic-controlanimals,theTg-KOmiceexhibitedpoorerbalance/coordination,moreanxiety,pooreroverallmnemonicfunctionde-creasedYM-PA,butmarkedlysuperiorobjectrecognition/identicationmemorydecreasedPR-Finlatency.However,theRML-FT4measurewasslightlylowerinTg-KOmice,suggestingbetterworkingmemoryfunction,relativetothenontrans-genicgroup.171

PAGE 182

AllFourGroupsStatistically-signicantdiscriminabilityamongthefourgroupswasreturnedbythedirect-entrydiscriminantanalysisWilks'slambda=0.010,p=.0275,with47%overallaccuracy.IndividualsoftheTggroupwerethemostlikelytobecor-rectlyidentied67%,whileonly22%oftheTg-KOmiceweredetected.Similarly,thestepwise-forwardapproachdemonstratedsignicantand,moreover,superiordis-criminabilityWilks'slambda=0.227,p=.0000.OnlyfourbehavioralmeasureswereretainedbythemodelOF,EP-TO,YM-PA,andRME-T4,whichaccuratelyclassied56%ofanimalsbygroup.TheNT-KOmicewerethemostlikelytobecorrectlyidentied5%,whileonly29%ofNTmicewererecognized.7.4ResultsofDataMiningAnalyses:ComprehensiveTaskBatteryAcomparativesummaryoftheperformanceofdatamining-basedclassiersuti-lizingbehavioralmeasuresfromthecomprehensivetaskbatteryisprovidedinTable7.4,forreference.Thefththroughseventhcolumnsofthetablereporttheperfor-manceevaluationi.e.,accuracy,sensitivity,andspecicityforeachclassier,withrespecttodiscriminabilitybetweenpairsofgroups,aswellasamongallfourgroups.7.4.1DecisionTreeAnalysisNontransgenic-ControlNTvs.Transgenic-ControlTgGroupsThedecisiontreeidentiedasinglebehavioralmeasure,RME-T5,asprovidingsucientinformationbiastodistinguishbetweenNTandTgmice.Theresult-ingclassiercorrectlyidentied74%ofanimalsbygroupKappa=0.45,with75%sensitivityand71%specicity,withrespecttotransgenicity.Alzheimer'stransgenic-controlmiceexhibitedsignicantcognitiveimpairmenti.e.,increasederrorsinover-allRAWMtrialT5,relativetonontransgenic-controlanimals,asindicatedinTable172

PAGE 183

7.1.ThisresultunderscorestheprimacyoftheRAWMtaskfordetectingcognitiveimpairmentand,specically,overallmeasuresofworkingmemory.Nontransgenic-ControlNTvs.Nontransgenic-KnockoutNT-KOGroupsDecisiontree-basedclassierswereunabletodistinguishbetweennontransgenicmicewiththeGRK5-wildtypeandanimalshavingtheGRK5-knockoutgenotype.AsreportedinTable7.4,noneofthebehavioralmeasuresofthecomprehensivetaskbatteryprovidedsucientinformationbiastoreliablydiscriminatebetweenthetwogroups.Thissuggestscomparablebehavioralperformancebetweenthetwogroups,i.e.,theabsenceofGRK5expressioninanontransgenicmousedoesnotsignicantlyimpacttheoverallbehavioralphenotypeofthemouse.Transgenic-ControlTgvs.Transgenic-KnockoutTg-KOGroupsTheRME-T5andRME-FT4measureswereselectedbythedecisiontreefortheirinformationvalueindistinguishingbetweenAlzheimer'stransgenicmicehavingtheGRK5-wildtypeandanimalswiththeGRK5-knockoutgenotype.Theclassiergen-eratedbythedecisiontreeexhibited71%overallaccuracyKappa=0.42,67%sensitivityand75%specicity,withrespecttotheGRK5-knockoutgroup.AlthoughtheclassierexhibitedsuperioraccuracyforidentifyingTgmice,relativetoTg-KOanimals,thechoiceoftworobustcognitivemeasuresofRAWMworkingmemoryindeed,bothnal-blockT4errorsandoverallT5errorshighlightstheimportanceoftheRAWMtaskforevaluatingcognitiveimpairment.Interestingly,althoughthesetwogroupsdiersignicantlyinonlyoneofthesetwomeasuresRME-T5,asshowninTable7.1,bothmeasurestogetherprovideremarkablediscriminativepo-tential.Alzheimer'stransgenic-controlmiceTgexhibitsignicantworkingmemorydecits,relativetotransgenicswhichdonotexpressGRK5.ThisresultssuggeststhatexpressionofGRK5,superimposedagainstanAlzheimer'stransgenicback-173

PAGE 184

ground,mayameliorateworkingmemorydysfunctionassociatedwiththeAPPswgenotype.Nontransgenic-ControlNTvs.Transgenic-KnockoutTg-KOGroupsAsshowninTable7.4,individualsfromthenontransgenicandtransgenic-knockoutgroupscouldnotbedistinguishedusinganycombinationofbehavioralmeasuresfromthecomprehensivetaskbatterybydecisiontree-basedclassiers.Noindividualmea-sure,orcombinationofmeasures,providedsucientinformationbiastoreliablysplitthedatabaseintothetwogroups.ThesendingssuggestthattheabsenceofGRK5expressioninanAlzheimer'stransgenicmousemaysucientlymitigatethebehavioralmanifestationsofAlzheimer's-likeneuropathology,astorendertheseindi-vidualscomparabletonontransgenicanimals,usingthecomprehensivetaskbatteryassessment.AllFourGroupsEightmeasuresfromthecomprehensivetaskbattery,representingadiversesam-plingofthemousebehavioralrepertoire,wereidentiedbythedecisiontreefordis-tinguishingamongthefourgroupsofanimals.Thesemeasureswere:OF,RME-FT4,RME-FT5,WM-Avg,EP-OE,YM-PA,BB,andRML-FT4.Thiscognitively-biasedsubsetofmeasuresaccuratelyidentied36%ofallanimalsbygroupKappa=0.14,asindicatedinTable7.4.7.4.2NeuralNetworkAnalysisNontransgenic-ControlNTvs.Transgenic-ControlTgGroupsTheneuralnetwork-basedclassierperformedremarkablywell,accuratelyiden-tifying89%ofanimalsbygroupKappa=0.77.Thesensitivityoftheclassierwas92%andthespecicitywas86%,withrespecttotransgenicity.Thislevelofperfor-174

PAGE 185

manceunderscorestherelativeeaseofidentifyingimpairedcognitiveperformanceintheAlzheimer'stransgenicanimals,relativetothenontransgenicmice.Nontransgenic-ControlNTvs.Nontransgenic-KnockoutNT-KOGroupsAsshowninTable7.4,onlytwo-thirdsofallnontransgenicanimalsweredistin-guishableonthebasisofGRK5genotypeKappa=0.34.AnimalswiththeGRK5-knockoutgenotypeweremorelikelytobemisclassied,relativetomicehavingtheGRK5-wildtype%sensitivityvs.71%specicity.Althoughthesetwogroupsdonotdiersignicantlywithrespecttoindividualcognitivemeasures,nontransgenicmiceexpressingGRK5weresignicantlymoreactivethanwereGRK5-knockoutan-imals,asmeasuredbytheOpenFieldtaski.e.,NT-KOmicedisplayedsignicantlyfewerline-crossingsintheopenarena.Transgenic-ControlTgvs.Transgenic-KnockoutTg-KOGroupsTheperformanceofneuralnetwork-basedclassiersfordistinguishingbetweenAlzheimer'stransgenicmiceonthebasisofGRK5expressionwasonlyslightlysu-periortorandom-chanceassignment52%overallaccuracy,Kappa=0.03.Indeed,detectionofTg-KOanimalsfailedtoreachcriterion0%.Nontransgenic-ControlNTvs.Transgenic-KnockoutTg-KOGroupsThelevelofdiscriminabilitybetweennontransgenicanimalsandmicehavingbothgeneticmanipulationsAlzheimer'stransgenicwithGRK5-knockoutwasonly56%overallKappa=0.15,asreportedinTable7.4.ThisperformancewaslargelydrivenbyaccuracyinidentifyingtheNTmicespecicity=71%,becauseaninsucientproportionofTg-KOanimalswereidentiedbytheclassiertoreachcriterion.AllFourGroupsThirty-ninepercentofallanimalswerecorrectlyassignedtotheirrespectivegroupsbytheneuralnetwork-basedclassierusingbehavioralmeasuresfromthecomprehensivetaskbatteryKappa=0.17.175

PAGE 186

7.4.3SupportVectorMachineAnalysisNontransgenic-ControlNTvs.Transgenic-ControlTgGroupsAsupportvectormachine-basedclassiertrainedusingallthebehavioralmea-suresfromthecomprehensivetaskbatterysuccessfullydistinguishedbetweennon-transgenicandAlzheimer'stransgenicmice,with74%overallaccuracyKappa=0.42.Thesensitivitywas83%andspecicitywas57%,asshowninTable7.4.Theobservedlevelofdiscriminabilitybetweenthesetwogroupsisnotsurprising,becauseAlzheimer'stransgenicmicediersignicantlyfromnontransgenicsinbothsensorimotorandcognitivemeasures.ComparedwithNTmice,Tganimalsexploretheopen-armsoftheY-mazemorefrequently,butexhibitpoorersystematicsearchoftheY-mazerelativelyfewerarm-visitalternations;YM-PA.TheAlzheimer'stransgenicmicealsodisplaypoorerworkingmemory,relativetonontransgenics,asindicatedbyhighererror-scoresRME-T4,andbothRME-T5andRME-FT5andlongerresponse-latenciesbothRML-T4andRML-T5intheRAWMtrialsT4andT5.Nontransgenic-ControlNTvs.Nontransgenic-KnockoutNT-KOGroupsThesupportvectormachine-basedclassierobtainedonly53%overallaccuracyKappa=0.07indistinguishingbetweennontransgenicanimalsonthebasisofGRK5expression.Classier-basedassignmentofNT-KOmicewascomparabletorandom-chanceperformance,however.Transgenic-ControlTgvs.Transgenic-KnockoutTg-KOGroupsAsindicatedinTable7.7,sixty-twopercentofAlzheimer'stransgenicmicewerecorrectlyidentiedonthebasisofGRK5-genotype.Although75%ofTganimalsex-pressingGRK5wererecognized,fewerthanhalfofthemicelackingGRK5expressionweresuccessfullydetected.176

PAGE 187

Nontransgenic-ControlNTvs.Transgenic-KnockoutTg-KOGroupsSupportvectormachine-basedclassiersfailedtodistinguishbetweenNTandTg-KOanimalsusingthebehavioralmeasuresfromthecomprehensivetaskbattery.Classierperformanceneverexceededtheaccuracyexpectedbyrandomassignmentofindividualstogroups.AllFourGroupsAsreportedinTable7.4,only28%ofallanimalswereaccuratelyidentiedbygroupbysupportvectormachinesKappa=0.03.Thislevelofperformanceonlyslightlyexceedstheaccuracyexpectedbyrandomassignmentofindividualstogroups.7.5ResultsofStatisticalAnalyses:InterferenceParadigm7.5.1StandardBehavioralAnalysisThebehavioralperformanceofallfourgroupsineachmeasureoftheinterferenceparadigmbotherror-scoresandresponse-latenciesareshowninFigure7.1,onthenextpage,withgroupmeansandassociatedstandarderrorsdepictedforeachmea-sure.BaseduponstandardstatisticaltestsANOVA,F-test,theTggroupdieredsignicantlyfromtheNTgroup,denotedbyanasterisk,withrespecttobotherrorsandlatenciesinthethree-trialrecall,retroactiveinterference,anddelayed-recalltasksallp.05.NosignicantdierenceswerefoundbetweentheNTandNT-KOgroups,withrespecttoanyofthebehavioralmeasuresintheinterferenceparadigm.Similarly,althoughtheTgandTg-KOgroupsdieredwithrespecttothethree-trialrecallerror-scorep=.075,thiscontrastdidnotmeetthesigni-cancecriterionp.05.Finally,NTandTg-KOanimalsexhibitsignicantp.05groupwisedierencesincognitiveperformance,asrevealedbytheretroactive177

PAGE 188

Figure7.1.GroupwisecontrastsforallinterferenceparadigmmeasuresintheGRK5study178

PAGE 189

interferencetasklatencies,butnoterrors,andindicatedwithapound-signsymbol#.Thisndingsuggeststhatnormali.e.,comparabletonontransgenicanimalscognitiveperformancewasobservedinAlzheimer'stransgenicanimalswhichdonotexpressGRK5inalltasks/measuresexceptretroactiveinterference,whereintheTg-KOmiceshowedimpairedcontext-switchingfromtheinterference-contextPoolBbacktotheoriginallearningconditionPoolA,relativetonontransgenics.More-over,itwastheresponse-latencymeasureoftheretroactiveinterferencetaskwhichdemonstratedgreatersensitivitytotheTg-KObehavioralphenotype.7.5.2CorrelationAnalysisTable7.5.CorrelationsbetweenbehavioralmeasuresoftheinterferenceparadigmintheGRK5study TT-L .72 .001 PI-E PI-L .95 .000 RI-E .65 .68 .003 .001 RI-L .63 .75 .97 .004 .000 .000 DR-E .61 .73 .87 .85 .005 .000 .000 .000 DR-L .54 .77 .89 .93 .95 .017 .000 .000 .000 .000 TT-E TT-L PI-E PI-L RI-E RI-L DR-E Signicantp<.05correlationsobservedbetweenpairsofmeasures,forallgroups,areindicatedinTable7.5.Markedcellsinthetableincludeboththecorre-lationcoecientr-value,topandsignicancep-value,bottom.Theabbreviationsforthefourtasksofthemouse-basedinterferenceparadigmare:Three-trialrecall,179

PAGE 190

errorsandlatencyTT-E,TT-L;proactiveinterference,errorsandlatencyPI-E,PI-L;retroactiveinterference,errorsandlatencyRI-E,RI-L;and,delayed-recall,errorsandlatencyDR-E,DR-L.Therewerepositivecorrelationsbetweentheerror-scoremeasuresofthethree-trialrecallandbothretroactiveinterferenceanddelayed-recalltasks.Asimilarpat-ternofassociationwasobservedamongthecorrespondinglatencymeasures.ThisisconsistentwithacquisitionoftheinitiallearningcomponentPoolAbeingpredictiveoflaterperformanceinthesamespatialenvironmenti.e.,PoolA.Theerror-scoreandresponse-latencymeasuresoftheproactiveinterferencetaskweresignicantlyintercorrelated.However,neithermeasuresfromtheproactiveinterferencetaskwassignicantlycorrelatedwithanymeasuresfromtheotherthreetasks,suggestingin-dependenceofperformancebetweenthetwolearningcontextsPoolAvs.PoolB.Signicantcorrelationswerefoundbetweencorrespondingerror-scoreandresponse-latencymeasuresfortheotherthreetasks,aswelle.g.,three-trialrecallerrorsandlatency,whichunderscoresthecomparabilityofthesetwoindicesofperformancewithintasksofthemouse-basedinterferenceparadigm.Additionalintercorrelationsacrosscognitivecomponents,aswellasbetweenerrorandlatencymeasures,weredetected.Bothmeasuresfromthethree-trialrecalltaskweresignicantlyintercorrelatedwithbothmeasuresofretroactiveinterference,sug-gestingconsistent/stableacquisitionandretentionoftheinitiallearningcomponentPoolA,despitetheinuenceofbothadistractortaskY-mazeexposureandtheinterferencetaskPoolBexposure.Similarly,completeintercorrelationbetweenallmeasuresoftheretroactiveinterferenceanddelayed-recalltaskssuggeststemporalstabilityoflearning.Thispermanenceisfurtherunderscoredbysignicantpairwiseintercorrelationamongallmeasuresofthethree-trialrecallanddelayed-recalltasks.180

PAGE 191

7.5.3FactorAnalysisTable7.6.Varimax-rotatedfactoranalysisofinterferenceparadigmmeasuresintheGRK5study Factor Measure I II RetroactiveInterferenceLatency 0.956 Delayed-RecallLatency 0.948 RetroactiveInterferenceErrors 0.944 Delayed-RecallErrors 0.934 Three-TrialRecallLatency 0.855 Three-TrialRecallErrors 0.744 ProactiveInterferenceErrors 0.992 ProactiveInterferenceLatency 0.981 Variance 60.99% 24.61% Avarimax-rotatedprincipalcomponentanalysisoftheeightmeasuresisshowninTable7.6.Signicantabsolutevaluegreaterthan0.700componentloadingsareindicatedforthetwofactorsreturned.Additionally,thecalculatedeigenvalueresultsi.e.,>1criterionwereinagreementwithscreeplotCattell,1966identi-cationoftwosignicantfactors.Theprimaryfactoraccountingforapproximately60%ofoverallvarianceconsistedoferror-scoreandresponse-latencymeasuresfromtheThree-TrialRecall,RetroactiveInterference,andDelayed-Recalltasks,andrep-resentsspatiallearningandmemoryfunctionassociatedwiththeinitiallearningconditionPoolA.Bycontrast,thesecondfactorapproximately25%ofvarianceincludesonlytheProactiveInterferencemeasures,andlikelyrepresentsspatiallearn-ingandmemoryrelatedtotheinterferenceconditionPoolB.181

PAGE 192

Table7.7.ClassierperformancecomparisonusinginterferenceparadigmmeasuresintheGRK5study Evaluation DiscriminantAnalysis Decision Neural Groups Dataset Criterion Complete Step-Fwd Tree Network SVM Accuracy NS 74% 63% 68% 53% Errors Sensitivity NS 67% 67% 75% 75% Specicity NS 86% 57% 57% 14% NT Accuracy NS 79% 68% 53% NS vs. Latency Sensitivity NS 75% 67% 58% NS Tg Specicity NS 86% 71% NS NS Accuracy NS 74% 58% 63% 63% Err+Lat Sensitivity NS 67% 58% 75% 67% Specicity NS 86% 57% 57% 57% Accuracy NS NS NS NS NS Errors Sensitivity NS NS NS NS NS Specicity NS NS NS NS NS NT Accuracy NS NS NS NS NS vs. Latency Sensitivity NS NS NS NS NS NT-KO Specicity NS NS NS NS NS Accuracy NS NS NS NS NS Err+Lat Sensitivity NS NS NS NS NS Specicity NS NS NS NS NS Accuracy NS NS 73% NS NS Errors Sensitivity NS NS 80% NS NS Specicity NS NS 67% NS NS Tg Accuracy NS NS NS NS NS vs. Latency Sensitivity NS NS NS NS NS Tg-KO Specicity NS NS NS NS NS Accuracy NS NS 55% NS NS Err+Lat Sensitivity NS NS NS NS NS Specicity NS NS 58% NS NS Accuracy NS 77% 53% 53% 53% Errors Sensitivity NS 80% 60% NS 90% Specicity NS 71% NS 58% NS NT Accuracy NS 88% 88% 71% NS vs. Latency Sensitivity NS 80% 100% 70% NS Tg-KO Specicity NS 100% 71% 71% NS Accuracy NS 88% 88% 65% NS Err+Lat Sensitivity NS 90% 100% 60% NS Specicity NS 86% 71% 71% NS Errors Accuracy 30% 32% 35% 30% 27% Allfour Latency Accuracy NS 46% 32% NS NS Err+Lat Accuracy NS 32% 43% 30% NS 182

PAGE 193

7.5.4DiscriminantAnalysisAsummarytablecomparingclassierperformanceisshowninTable7.7.Thefourthandfthcolumnsofthetabledisplaytheresultsofdiscriminantanalysis-basedclassiersusingthedirect-entrycompleteandstepwise-forwardmethods,respec-tively.Measuresofclassierperformancei.e.,accuracy,sensitivity,andspecicityaredepictedonlywhentheobservedlevelstrictlyexceedsexpectedvaluesassociatedwithrandom-chanceassignmentofindividualstogroupsi.e.,50%fortwo-groupdiscriminability,25%forfour-groupdiscriminability.Nontransgenic-ControlNTvs.Transgenic-ControlTgGroupsAdirect-entrydiscriminantanalysisdidnotreturnsignicantdiscriminabilitybe-tweenthetwogroupsWilks'lambda=0.585,p=.0913usingonlyerror-scoredata.Bycontrast,astepwise-forwardanalysisoferror-scoreswassuccessfulWilkslambda=0.643,p=.0069with74%overallaccuracysensitivity=67%,specicity=86%,asshowninTable7.7,whereinonlythedelayed-recallerrormeasurewasretainedbythestepwise-forwardmodel.Direct-entrydiscriminantanalysisusingonlyresponse-latencydatafailedtodistinguishbetweenthetwogroups,althoughstepwise-forwardanalysisreturnedsignicantdiscriminabilityWilkslambda=0.677,p=.0111with79%accuracysensitivity=75%,specicity=86%,retainingonlythethree-trialrecalllatencymeasureasthepredictorvariable.Whenbotherror-scoresandresponse-latencieswereincluded,onlythestepwise-forwardanalysissuccessfullydis-tinguishedindividualsbytransgenicityWilkslambda=0.643,p=.0069with74%accuracysensitivity=67%,specicity=86%,asdepictedinTable7.7.Thedelayed-recallerror-scorewastheonlybehavioralmeasureoftheeightprovidedtoberetainedbythestepwise-forwardmodel.Hence,stepwise-forward,butnotdirect-entry,discriminantanalysessuccessfullydistinguishedbetweennontransgenic-183

PAGE 194

controlandtransgenic-controlmice,utilizingeitherthedelayed-recallerror-scoreorthethree-trialrecalllatencymeasure.Betweenthesetwomeasures,however,theformerwasshowntobesuperiorwhenallmeasureswereavailableforselection.Ani-malsintheTggroupexhibitedhigheri.e.,poorervaluesforbothofthesemeasures,relativetomiceintheNTgroup.Nontransgenic-ControlNTvs.Nontransgenic-KnockoutNT-KOGroupsNeitherdirect-entrynorstepwise-forwarddiscriminantanalysesshowedsigni-cantdiscriminabilitybetweenthetwogroups,usingeithererror-scoresorresponse-latencies,orboth.Thestepwise-forwardanalysisindicatedthatnoneofthecan-didatepredictorvariablesmetthecriterionformodelinclusionalpha-to-enter=0.15.Thus,removalofGRK5expressionhadnoeectonbehavioralperformanceinnormalnontransgenicmice.Transgenic-ControlTgvs.Transgenic-KnockoutTg-KOGroupsBothdirect-entryandstepwise-forwarddiscriminantanalysesfailedtodistin-guishbetweenthetwogroupsusingerror-scores,response-latencies,orbothbehav-ioralmeasures.Noneofthecandidatepredictorvariablesmetthestatisticalcri-terionalpha-to-enter=0.15tobeaddedtothestepwise-forwardmodel.SimilartotheresultsobtainedfornontransgenicanimalsGRK5-wildtypevs.knockout,neitherdiscriminantanalysis-basedclassierdistinguishedbetweenGRK5-wildtypeandknockoutgenotypeinAlzheimer'stransgenicmice.Hence,removalofGRK5expressiondidnotenhancethebehavioralimpairmentofTgmice.Nontransgenic-ControlNTvs.Transgenic-KnockoutTg-KOGroupsThedirect-entrydiscriminantanalysiswasunabletodistinguishsignicantlybe-tweenthetwogroupsusingerror-scoresand/orresponse-latencies.Bycontrast,thestepwise-forwardmethodreturnedsignicantgroupwisediscriminabilityforallthreedatasetconditions.Usingonlyerror-scores,thestepwise-forwardclassierretained184

PAGE 195

theproactiveinterferenceandretroactiveinterferencemeasures,anddemonstrated77%overallaccuracyWilks'slambda=0.627,p=.0379with80%sensitivityand71%specicity.Similarly,whenresponse-latencieswereused,onlytheproac-tiveinterferenceandretroactiveinterferencemeasureswereretainedbythemodel,exhibiting88%accuracyWilks'slambda=0.606,p=.0299with80%sensitiv-ityand100%specicity;allnontransgenicanimalswerecorrectlyidentiedbythestepwise-forwardclassierusinglatencydataalone.Whenbotherror-scoresandresponse-latencieswereprovided,thestepwise-forwardanalysisretainedtheproac-tiveinterferenceerror-scoreandtheretroactiveinterferencelatencymeasures,andshowed88%accuracyWilks'slambda=0.549,p=.0151,90%sensitivityand86%specicity.Hence,theNTandTg-KOgroupsweremoreeasilydistinguishablethanNTvs.Tg,suggestingadditionalcognitiveimpairment,beyondtheAPPtransgeniceect,maybeattributedtotheabsenceofGRK5expression.AllFourGroupsDirect-entrydiscriminantanalysisreportedsignicantdiscriminabilityamongthefourgroupsWilks'lambda=0.515,p=.0475,usingonlyerror-scoredata.How-ever,theoverallclassicationaccuracywas30%.Astepwise-forwardanalysisalsoreturnedsignicantdiscriminabilityWilkslambda=0.675,p=.0043withmodest32%overallaccuracy,asshowninTable7.7.Onlythedelayed-recallerrormea-surewasretainedbythestepwise-forwardmodel.Bycontrast,usingonlyresponse-latencydata,onlythestepwise-forwarddiscriminantanalysissignicantlydistin-guishedamongthefourgroupsWilkslambda=0.681,p=.0049withmoderate46%accuracy,asreportedinTable7.7.Theonlypredictorvariableincludedinthestepwise-forwardmodelwasthethree-trialrecalllatencymeasure.Similarly,whenbotherrorsandlatencieswereprovided,onlythestepwise-forwardapproachreturnedsignicantdiscriminabilityWilkslambda=0.675,p=.0043with32%accuracy,185

PAGE 196

asportrayedinTable7.7.Thedelayed-recallerrormeasurewastheonlypredictorvariableretainedbythemodelfordistinguishingamongthefourgroups.Hence,thedelayed-recallerrormeasurewasidentiedasthebestdiscriminatoramongthefourgroupsofmice,althoughthethree-trialrecalllatencymeasurealsodemonstratedpotentialfordistinguishingamongthegroups.Itshouldbenoted,however,thattheoveralllevelsofaccuracyobservedi.e.,30%to46%areonlymodestlysuperiortotherandom-chanceperformancecriterionforfour-groupdiscriminability%.7.6ResultsofDataMiningAnalyses:InterferenceParadigmTable7.7comparestheperformanceofclassiersbasedonadvancedstatisticaltechniquescolumnsfourandveanddatamining-basedmethodscolumnssixthrougheight.Performancemeasuresi.e.,accuracy,sensitivity,andspecicityarereportedonlywhenthecriterionforrandom-chanceassignmenti.e.,50%fortwo-groupdiscriminability,25%forfour-groupdiscriminabilityisexceeded.7.6.1DecisionTreeAnalysisNontransgenic-ControlNTvs.Transgenic-ControlTgGroupsUsingonlyerror-scoredata,thedecisiontreeattemptedtosplitthedatasetusingthreemeasuresdelayed-recall,proactiveinterference,andthree-trialrecall,result-ingin63%correctclassicationofcasesKappa=0.23,asshowninTable7.7.Thesensitivitywas67%andspecicitywas57%,withrespecttotransgenicity.Whenonlyresponse-latencydatawereprovided,thedecisiontree-basedclassierusedtwomeasuresthree-trialrecallandproactiveinterferencetodistinguishbetweenthetwogroups,resultingin68%overallaccuracyKappa=0.36.Forthelatency-basedclassication,theclassiersensitivitywas67%andspecicitywas71%.Includingbotherror-scoresandresponse-latenciesresultedinadecisiontree-basedclassier186

PAGE 197

withthreemeasuresdelayed-recallerrors,proactiveinterferenceerrors,andthree-trialrecallerrorsselectedtodistinguishbetweenthetwogroups,anddisplaying58%correctclassicationofcasesKappa=0.15.Thesensitivitywas58%andspecicitywas57%,asindicatedinTable7.7.Theseanalysessuggestthatmeasuresfromthethree-trialrecallandproactiveinterferencetasksmostreliablydistinguishanimalsbytransgenicityAlzheimer'stransgenicvs.nontransgenic,andthedelayed-recallerrormeasurealsocontributestothediscriminability.However,thelevelofdiscrim-inabilityi.e.,58%to68%observedwasonlyslightlybetterthanrandom-chanceassignmentexpectationfortwo-groupclassication0%.Nontransgenic-ControlNTvs.Nontransgenic-KnockoutNT-KOGroupsDecisiontree-basedclassierswereunabletodistinguishbetweenGRK5-wildtypeandknockoutgenotypenontransgenicanimalsusingeithererror-score,response-latency,orbothsetsofbehavioralmeasures.Noneofthecandidatepredictorvari-ablesmettheinformationgaincriterionrequiredbythedecisiontreesforsplittingthedatasetintogroups.Transgenic-ControlTgvs.Transgenic-KnockoutTg-KOGroupsUsingonlytheerror-scoremeasures,decisiontree-basedclassierswereabletodistinguishbetweenthetwogroupsusingonlythedelayed-recallmeasure,resultingin73%overallaccuracyKappa=0.46.Thesensitivitywas80%andspecicitywas67%,asindicatedinTable7.7.However,theclassierwasunabletogenerateadeci-siontreetodistinguishbetweenthetwogroupsusingonlyresponse-latencies.Whenprovidedwithbotherror-scoresandreponse-latencies,thedecisiontree-basedclassi-erselectedtwomeasuresdelayed-recallerrorsandretroactiveinterferencelatencytodistinguishbetweenthetwogroups,resultingin55%correctclassicationofcasesKappa=0.08,asreportedinTable7.7.Theclassier'ssensitivitywasonly50%comparabletorandom-chanceassignmentandspecicitywas58%.Takentogether,187

PAGE 198

theseresultssuggestthesignicantdiscriminabilityobservedforerrors+latenciesisdrivenbytheerror-scoredata,since:Response-latencymeasuresalonedonotprovidediscriminability,andthelevelofdiscriminabilityaccuracywasmuchhigherforerror-scoresalone,relativetoerrors+latenciescombined.Nontransgenic-ControlNTvs.Transgenic-KnockoutTg-KOGroupsThedecisiontree-basedclassierwasonlymodestlycapableofdistinguishingbe-tweenthetwogroupsusingerror-scoresalone,displayingonly53%accuracyKappa=0.03with60%sensitivity,utilizingtheproactiveinterferenceandretroactivein-terferencemeasures.Bycontrast,88%accuracyKappa=0.75wasachievedusingonlyresponse-latencies,whereinonlytheretroactiveinterferencemeasurewasse-lectedforitsinformation-biascapacity.AlloftheTg-KOanimalswerecorrectlyidentied0%sensitivityusinglatencydataalone.Similarly,whenbotherror-scoreandresponse-latencymeasureswereavailable,theoverallaccuracywas88%Kappa=0.75,with100%sensitivityand71%specicity,usingonlytheretroactiveinterferenceresponse-latencymeasuretosplitthedataset.Hence,groupwisedier-encesinasinglebehavioralmeasureretroactiveinterferencelatencyaresucienttodistinguishbetweenNTandTg-KOanimals.AllFourGroupsThedecisiontreeattemptedtodistinguishamongthefourgroupsusingonlyerror-scoredata,andidentiedtwobehavioralmeasuresretroactiveinterferenceandthree-trialrecallwhichoptimallyclassiedindividualsintotheirrespectivegroups.AsshowninTable7.7,35%correctclassicationofcasesKappa=0.13wasob-served.Whenonlyresponse-latencydatawereprovided,thedecisiontreeaccuratelydistinguished32%oftheindividualanimalsKappa=0.07usingallfourlatencymeasures.Noneofthenontransgenic-controlNTanimalswerecorrectlyidenti-edusingonlyerror-scoreorresponse-latencydata,however.Providingthedecision188

PAGE 199

treewithbotherror-scoreandresponse-latencydataforthefourgroupsresultedinselectionoffourmeasuresretroactiveinterferenceerrors,three-trialrecallerrors,delayed-recallerrors,andretroactiveinterferencelatency,andoverallaccuracyof43%Kappa=0.24,asindicatedinTable7.7.Hence,discriminabilityamonggroupsbydecisiontree-basedclassiersappearssignicantstatistically,butnotappreciablyaboverandom-chancelevelsi.e.,25%.7.6.2NeuralNetworkAnalysisNontransgenic-ControlNTvs.Transgenic-ControlTgGroupsWheneithererror-scoresorresponse-latencieswereused,theoptimalarchitectureoftheneuralnetwork-basedclassierconsistedoffourinput-layercomputingunitsthebehavioralmeasures,fourhidden-layerunits,andtwooutput-layerunitsthetwogroups.Forerrors-only,theclassiershowed68%overallaccuracyKappa=0.32,with75%sensitivityand57%specicity,withrespecttotransgenicity,asindicatedinTable7.7.Usingonlylatencydata,theoverallaccuracywas53%Kappa=0.01,sensitivitywas58%andspecicitywas43%.Hence,error-scoresdistinguishbetweennontransgenicandAlzheimer'stransgenicanimalsmoreaccuratelythandoresponse-latencies;threeoutoffourtransgenicmicewerecorrectlyidentied.Usingbotherrorsandlatencies,theoptimalneuralnetworkarchitectureincludedeightinput-layercomputingelementsthebehavioralmeasures,fourhidden-layerunits,andtwooutput-layerunitsthetwogroups.Theresultingclassiercorrectlyidentied63%ofcasesanimalsbygroupKappa=0.23.Thesensitivitywas75%andspecicitywas57%.Theinclusionoflatencymeasureswiththeerror-scoredata,therefore,underminedoveralldiscriminabilitybetweenthegroups.189

PAGE 200

Nontransgenic-ControlNTvs.Nontransgenic-KnockoutNT-KOGroupsNoneoftheneuralnetwork-basedclassierssuccessfullydistinguishedbetweenGRK5-wildtypeandknockoutgenotypeinnontransgenicmiceusingeithererror-scores,response-latencies,orboth.Theresultingneuralnetworksdidnotreliablyclassifyindividualsbygroup,basedonresultsfrommultipleprogramexecutions.Transgenic-ControlTgvs.Transgenic-KnockoutTg-KOGroupsNeuralnetwork-basedclassierstrainedusingerror-scores,response-latencies,orboth,failedtodistinguishreliablybetweenGRK5-wildtypeandknockoutgenotype,superimposedonanAlzheimer'stransgenicmousebackground.Indeed,signicantclassierperformanceinstabilitywasobservedacrossrepreatedprogramruns.Nontransgenic-ControlNTvs.Transgenic-KnockoutTg-KOGroupsTheneuralnetworkshowedonlymarginalperformance,relativetorandom-chanceassignment,withonly53%ofanimalscorrectlyidentiedKappa=0.07usingonlyerror-scoremeasures.Indeed,only58%ofnontransgenicswerecorrectlyidentiedspecicity,whilethesensitivitywasnonsignicant.Latencymeasuresalone,bycontrast,accuratelydistinguished71%oftheanimalsbygroupKappa=0.41,with70%sensitivityand71%specicity.Whenbotherror-scoresandresponse-latencieswereprovided,however,theperformancedeclinedto65%accu-racyKappa=0.30,withonly60%sensitivityand71%specicity.Theseresultssuggestthatresponse-latenciesdrivediscriminabilitybetweenthetwogroups,withrespecttoneuralnetwork-basedclassiers.AllFourGroupsUsingonlyerror-scoredata,theoptimalneuralnetwork-basedclassierconsistedoffourinput-layercomputingunitsthebehavioralmeasures,threehidden-layerunits,andfouroutput-layerunitsthefourgroups.Theclassierdemonstrated30%overallaccuracyKappa=0.05,asreportedinTable7.7,althoughnoneof190

PAGE 201

thecontrolanimalsNTorTgwerecorrectlyidentied.Classierstrainedusingonlyresponse-latencydatadidnotsuccessfullydistinguishamongthefourgroups.Theoptimalclassierutilizingbotherror-scoresandresponse-latencies,however,wascomprisedofeightinput-layercomputingelementsthebehavioralmeasures,threehidden-layerunits,andfouroutput-layerunitsthefourgroups,andcorrectlyidentied30%ofcasesanimalsbygroupKappa=0.05.Interestingly,noneofthenontransgenicanimalsNTandNT-KOwerecorrectlyidentied.Hence,theinclusionoflatencymeasureswitherrormeasuresneitherimprovednorcompromisedoveralldiscriminabilityamongthefourgroups,relativetoexclusiveuseoferror-scoresaspredictorvariables.Moreover,theobserveddiscriminabilityamongthefourgroups0%wasonlyslightlybetterthanclassicationperformanceexpectedbyrandom-assignment5%.7.6.3SupportVectorMachineAnalysisNontransgenic-ControlNTvs.Transgenic-ControlTgGroupsThesupportvectormachine-basedclassierobtainedonly53%overallaccuracyKappa=-0.12usingonlyerror-scoremeasures,with75%sensitivityand14%speci-city,asshowninTable7.7.MostanimalswereclassiedasTg,regardlessofactualgenotype.Bycontrast,thesupportvectormachinewasunabletodistinguishbe-tweengroupssolelyonthebasisofresponse-latencymeasures;allNTanimalsweremisclassiedastransgenic.Byincludingbotherror-scoresandresponse-latencies,however,theclassierdemonstrated63%overallaccuracyKappa=0.23,asindi-catedinTable7.7.Thesensitivitywas67%andspecicitywas57%.Nontransgenic-ControlNTvs.Nontransgenic-KnockoutNT-KOGroupsSupportvectormachine-basedclassierswereunabletodistinguishbetweenGRK5-wildtypeandknockoutgenotypenontransgenicanimalsusingbehavioralerror-scores,191

PAGE 202

response-latencies,orboth.Classierperformanceneverexceededtheaccuracyex-pectedbyrandomassignmentofindividualstogroupsi.e.,50%.Transgenic-ControlTgvs.Transgenic-KnockoutTg-KOGroupsAlzheimer'stransgenicanimals,withorwithouttheGRK5-knockoutgenotype,werenotdistinguishablebysupportvectormachine-basedclassiersusingeithererror-scores,response-latencies,orboth.Theperformanceaccuracyneverexceeded50%,theexpectedvaluebasedonrandomassignmentofindividuals.Nontransgenic-ControlNTvs.Transgenic-KnockoutTg-KOGroupsOnlythesupportvectormachine-basedclassiertrainedusingerror-scoremea-suresalonewasabletodistinguishbetweenthesetwogroups,andperformedonlyslightlybetterthanrandom-chanceassignment,with53%accuracyKappa=-0.12.Indeed,although90%oftheTg-KOanimalswerecorrectlyidentiedsensitivity,alloftheNTanimalswereincorrectlyidentiedasTg-KO.Hence,thisclassierperformedsimilarlytotheNTvs.Tgclassier,withrespecttodiscriminabilityandassignmentprole.AllFourGroupsUsingonlyerror-scoresfortraining,theclassiercorrectlyidentied27%ofallanimalsbygroupKappa=0.001,asreportedinTable7.7.Transgenic-controlTgmicewerethemostlikelytobecorrectlyidentiedsensitivity=67%,how-evernoneofthenontransgenic-controlNTortransgenic-knockoutTg-KOwererecognizedbytheclassier.Bycontrast,theclassierwasunabletodistinguishamongthefourgroupsusingonlyresponse-latencydata.Whenbotherror-scoresandresponse-latencieswereusedfortrainingtheclassier,theoveralldiscriminabil-ityamonggroupswasnotsignicanti.e.,lessthanexpectedlevelattributedtorandomassignment.Hence,theinclusionofresponse-latencydataunderminedthemodestpotentialoferror-scoremeasuresfordistinguishingamongthemicebygroup.192

PAGE 203

Reminiscentofthefour-groupclassicationperformanceofdecisiontreesandneuralnetworks,theobserveddiscriminabilitywasonlymarginallysuperiortorandom-chanceassignmentusingerror-scoresalone.7.7DiscussionThecognitivedecitsnotedbySuoetal.2007inagednontransgenicGRK5-knockoutmicewereconnedtospecicdomains.Indeed,basicmnemonicfunction,referencememoryandspatialreferencelearning,aswellasobjectidenticationabil-itywerecomparabletoage-matchedGRK5-wildtypeanimalsSuoetal.,2007.At17to19monthsofage,nodecitswereobservedintheY-mazepercentalterna-tions,Morriswatermazebothacquisitionandretention,circularplatform,andplatformrecognitiontasksSuoetal.,2007betweenGRK5-wildtypeandknockoutmice.However,short-termworkingmemoryimpairmentwasobservedinGRK5-knockoutanimalsi.e.,increasedRAWMT4andT5errors,relativetowildtypeSuoetal.,2007,basedonone-wayANOVAacrossdays.Inthepresentstudy,adierentgroupofGRK5-knockoutmicewasinvestigated,alongwithcombinedGRK5-knockoutandTg2576APPmice.Thepurposewastodetermineifanygreatercognitiveimpairmentwasevidentinthecombined-genotypemice,relativetoeitherGRK5-knockoutorTg2576alone.Allanimalscompletedboththecomprehensivebehavioraltaskbatteryandthenovelmouse-basedinterferenceparadigm.Asdiscussedearlier,neurobehavioralinvestigationsofgroup-dierencese.g.,transgenicity,therapeuticecacytypicallyutilizearel-ativelysmallsubsetofmultivariatestatisticalapproaches.Moreover,analysisofvarianceANOVA,F-testand/orregressionarefrequentlytheonlymethodscon-sidered,whilemore-appropriatetechniquesmaybediscountedor,indeed,ignored,despitethewidespreadavailabilityofpowerfulcomputingsoftwareandreferencedoc-193

PAGE 204

umentation.Forthisreason,thepresentstudyutilizedbothstandardandadvancedstatisticalmethodsforanalyzingthebehavioralmeasuresobtainedfrombothbehav-ioralassessmentregimensi.e.,comprehensivetaskbattery,mouse-basedinterferenceparadigm.Inadditiontostatisticalmethodologies,datamining-basedapproacheswerealsoutilizedfordetermininggroupwisediscriminability,bothbetweenpairsofgroupsandamongallgroupsofmice.Standard,ANOVA-basedgroupwisecomparisonswereperformedusingallmea-suresofthecomprehensivetaskbattery,toidentifysignicantpairwisedierencesinbehavioralmeasures.SignicantdierencesbetweentheNTandTg,NTandNT-KO,andTgandTg-KOgroupsareindicated,aswell.Nontransgenic-controlandAlzheimer'stransgenic-controlanimalswerefoundtodierinbothsensorimo-torEP-OEandcognitivemeasuresYM-PA,RAWMoverallT4andT5errorsandlatencies,andRAWMnal-blockT5errors.TheprimacyofRAWMmeasuresun-derscorestheimportanceofworkingmemoryassessmentforidentifyingAlzheimer's-associatedcognitiveimpairment,aswellastheutilityoftheRAWMtaskforbehav-ioralphenotypinginmice.OnlytwosignicantdierenceswerefoundbetweentheTgandTg-KOanimalsbothsensorimotor,OFandEP-OE,suggestingtheremovalofGRK5expressiondoesnotfurtherexacerbatecognitiveimpairmentobservedinAlzheimer'stransgenicanimals.Similarly,nontransgenicanimalsexpressingGRK5donotdiersignicantlyfromGRK5-knockoutmicewithrespecttoanycognitivemeasure,althoughthesetwogroupsdierinOpenFieldactivity.Hence,onthebasisofstandardstatisticalmethods,wemayconcludethatGRK5doesnotsignif-icantlyimpactcognitiveperformanceinmice,regardlessofAlzheimer'stransgenicbackground.However,nosignicantdierenceswerefoundforanyofthebehavioralmeasuresbetweentheNTandTg-KOgroupsofmice.TakentogetherwiththeNTvs.Tgcontrast,thisndingsuggeststhattheabsenceofGRK5expressionmay,in194

PAGE 205

fact,amelioratecognitiveimpairmentnormallyobservedinAlzheimer'stransgenicanimals.Thecorrelationanalysisresultswereconsistentwithpriorstudiese.g.,Leighty,etal.,2004;CaeineAdministrationinNontransgenicMicestudy,thisdisserta-tioninvolvingbehavioralassessmentinmiceusingthecomprehensivetaskbattery,whereinwidespreadsignicantcorrelationsarefoundamongcognitivetasksspeci-cally,PlatformRecognition,Morriswatermaze,andRAWMand,toalesserextent,betweensensorimotorandcognitivetaskse.g.,OpenFieldactivityandRAWM.Additionally,themultifactorialnatureofcertaintaskse.g.,ElevatedPlusmaze,Y-mazewasrevealedthroughsegregationofcomponentmeasurecorrelations,suchasbetweenexploratory/activityelementsoftheElevatedPlusmazeandmeasuresofgenerallocomotorfunctionEP-CE/OForsystematicsearchcapacityEP-OE/YM-PA.Theaggregationofcognitive-basedperformancemeasuresacrosstaskswasrevealedthroughexploratoryfactoranalysis,whereintheprimaryfactor,represent-inggeneralmnemonicfunction,wascomprisedofallRAWMmeasures,aswellasY-mazepercentalternations,whileaseparatefactorwascomprisedofbothPlatformRecognitionandMorriswatermazetaskmeasures.Segregationoftaskcomponents,aswell,wasfoundthroughfactoranalysise.g.,thethreeElevatedPlusmeasuresloadonseparatefactors.Alsoutilizingdatafromthecomprehensivetaskbattery,discriminantanalysisbothdirect-entryandstepwise-forwardwasusedtoconstructlinearclassiersfordistinguishingindividualsbetweengroups,aswellasamongallfourgroups.Sig-nicantdiscriminabilityamongthegroupswasobtainedusingstepwise-forwarddis-criminantanalysis,althoughthestandarddirect-entryapproachwasunsuccessful.Thesendingswereconsistentwithearlierstudies,whereinstepwise-forwardanaly-sesachievesuperiordiscriminabilityrelativetothestandarddirect-entrycomplete195

PAGE 206

approache.g.,ArendashandKing,2002;Leightyetal.,2004.Optimaldiscrim-inabilitywasobservedbetweenthenontransgenicanimalgroupsonthebasisofGRK5expression,inwhich87%ofmicewerecorrectlyidentiedusingonlythreeofthebehavioralmeasuresOF,EP-TO,andRME-FT5,representingacombinationofsensorimotor,anxiety,andcognitivecomponentsofbehavior.Inaddition,thestepwise-forwardanalysisdemonstratedthehighestoveralldiscriminabilityamongthefourgroupsofallclassiersexamined,correctlyidentifying56%ofindividuals,yetrequiringonlyfourofthecomprehensivetaskbatterymeasures.TheprimacyofcognitivemeasuresfromtheRAWMtask,emphasizedbybothstandardANOVAandadvancedstatisticalmethods,furtherunderscorestheimportanceofworkingmemoryevaluationinbehavioralphenotypingandclassication.Takentogether,theseresultsdemonstratetheremarkablecapacityofstepwise-forwarddiscriminantanalysisforidentifyinggroupwisedierences,oftenusingcombinationsofpredictorvariableswhichmaynotappearmeaningfulorcoherenttoanexperiencedanimalbehaviorist,forexample.Althoughaniterativesearchthrougharelativelylargearrayofempiricalmeasuresmaybeacceptableforexploratorydataanalysis,thisisnotnecessarilyappropriatewhenpriorexperienceorotherguidanceexistsforidentifyingcandidatepredictorvariables.Indeed,theinformedselectionofindivid-ualbehavioralmeasurese.g.,onlyRAWMerror-scoresfollowedbytheapplicationofstandardstatistics,includingANOVA,isacommonprotocolforneurobehavioralresearch,particularlywhenreliablemeasuresforspecicsyndromesareknowninadvancee.g.,useofRAWMtrialsT4andT5toassessworkingmemoryimpairmentassociatedwithAlzheimer's-likeneuropathology.Thedatamining-basedclassiersexhibiteddiverse,idiosyncraticbehavioracrossbothgroupsandmethodologies,incontrasttotherelativelyconsistentperformanceofthestepwise-forwardanalyses.Indeed,theonlyconsistentresultacrossalldata196

PAGE 207

miningtechniqueswasthediscriminabilitybetweentheNTandTggroups.Decisiontreesexhibitedthepoorestperformanceintermsofpairwisegroupdiscriminabil-ity,onlydistinguishingbetweentwopairsofgroupsNTvs.Tg,Tgvs.Tg-KO,whileneuralnetworksdemonstratedthebestoverallperformanceamongthedataminingtechniquesforbothtwo-wayandfour-wayclassication.Infact,theneu-ralnetwork-basedclassieroutperformedthestepwise-forwardanalysisfordistin-guishingbetweenNTandTganimals9%vs.84%.Neuralnetworks,however,performedonlyslightlybetterthanrandom-chancelevelwithrespecttotheTgvs.Tg-KOandNTvs.Tg-KOcomparisons.Thesupportvectormachine-basedclassi-ersshowedthepoorestfour-waydiscriminabilityofallclassiersexamined28%,relativetothe25%expectedthroughrandomassignmentofindividualstogroups.MeasuresfromtheRAWMwerepreferentiallyselectedbythedecisiontrees,under-scoringthediagnosticutilityandinformationrichnessofthiscognitiveassessmenttask.Hence,despitetheavailabilityofdiversebehavioralmeasuresfromthecom-prehensivetaskbattery,datamining-basedapproachesweregenerallyweakerthanadvancedstatisticalmethods,withrespecttogroupwisediscriminabilityinvolvingtwogeneticmanipulationsi.e.,APP,GRK5inmice.Incontrasttothecomprehensivetaskbattery,themouse-basedinterferenceparadigmconsistsonlyofcognitive-basedbehavioralmeasures.Botherror-scoresandresponse-latenciesarereportedforalltasksthree-trialrecall,proactiveinter-ference,retroactiveinterference,anddelayed-recall.StandardstatisticalanalysesANOVAindicatedsignicantp<.05dierencesbetweentheTgandNTgroupsinthethree-trialrecall,retroactiveinterference,anddelayed-recallmeasures,withrespecttobotherrorsandlatencies.Allthreeofthesemeasuresreectacquisi-tionandretentionoftheinitiallearningtaskPoolA,bycontrastwiththeproac-tiveinterferencetaskPoolBwhichrepresentsadistinct,albeitcognitively-related197

PAGE 208

spatialworkingmemory,learningcondition.Thiscontext-sensitivityeectwasunderscoredbypatternsofcorrelationandfactorcomponentloadingsobservedinsubsequentanalysesoftheinterferenceparadigmmeasures.NeithernontransgenicsnorAlzheimer'stransgenicmicecouldbedistinguishedonthebasisofGRK5ex-pressioni.e.,NTvs.NT-KO,Tgvs.Tg-KO,usinganyofthetasks/measuresoftheinterferenceparadigm.Furthermore,therewerenosignicantgroupwisedier-encesobservedineithererror-scoresorresponse-latenciesintheproactiveinterfer-encetask.Indeed,thecomparativelyhigherscoresintheproactiveinterferencetaskindicativeofpoorertaskperformance,relativetotheothertasksexhibitedbyallgroupsunderscoresthedicultyofswitchingbetweenPoolA-andPoolB-learningexperiencedbyallanimalstested.Additionally,theTg-KOmiceonlydieredsig-nicantlyfromNTmicewithrespecttoretroactiveinterferenceresponse-latency,whichinvolvescontext-switchingbetweentheinterference-conditionPoolBandtheoriginallearningconditionPoolA.Hence,generallycomparablecognitiveperfor-mancewasdisplayedbyNTandTg-KOanimals,consistentwithstandardstatisticalanalysisANOVA,butcontrarytoadvancedstatisticalexaminationdiscriminantfunctionofbehavioralmeasuresfromthecomprehensivetaskbatterywhereinaclas-siermodelusedsensorimotorandcognitivemeasurestoreliablydistinguishbetweenthesetwogroups.Thecorrelationanalysisofbehavioralmeasuresfromtheinterferenceparadigmrevealedsignicantintercorrelationamongbotherrorandlatencymeasuresofthethree-trialrecall,retroactiveinterference,anddelayed-recalltasks,reminiscentofcorrelationpatternsobservedbothwithinandbetweencognitive-basedtasksofthecomprehensivetaskbatterye.g.,Leightyetal.,2004.Inaddition,correspondingerrorandlatencymeasuresforeachtaskwerealsostronglycorrelated,underscoringthecomparabilityofthesecognitiveperformanceindices.Neitheroftheproactive198

PAGE 209

interferencemeasureswassignicantlycorrelatedwithanyotherbehavioralmeasure,whichsuggestsindependencebetweenlearningoccurringwithinproactiveinterfer-enceandtheothermeasuresevaluatedfromPoolA.Indeed,thetwosignicantfactorsreturnedbyexploratoryfactoranalysisclearlysegregatedmeasuresassociatedwithPoolAFactorI;three-trialrecall,retroactiveinterference,anddelayed-recallfromthemeasuresassociatedwithPoolBFactorII;proactiveinterference.Theresultsofthediscriminantanalysesofmeasuresfromthecognitiveinter-ferenceparadigmweregenerallyconsistentwithearlierstudies,whereinstepwise-forwardanalysesachievesuperiordiscriminabilityrelativetothestandarddirect-entrycompleteapproache.g.,ArendashandKing,2002;Leightyetal.,2004.In-deed,theonlysuccessfulstatistical"classierutilizedstepwise-forwarddiscriminantanalysisforbothTgvs.NT,andNTvs.Tg-KO,discriminabilityusingerror-scoresand/orresponse-latencydata.Alzheimer'stransgenicanimalswerenotdistinguish-ablewithrespecttoGRK5expression,norwereNTmicesignicantlydistinguishablefromNT-KOanimalsconsistentwithSuoetal.,2007.Moreover,comparativelysuperiordiscriminabilitybetweentheNTandTg-KOgroupswasdemonstratedus-ingthetwointerference-relatedmeasuresproactiveandretroactive,whiletheNTvs.Tgcontrastemphasizedthedelayed-recallerrorscore.Hence,thedierencesincognitiveperformancewhichdistinguishNTandTg-KOindividualsdierfromthosewhichdistinguishNTandTgindividuals.Indeed,theseresultswouldbepredictablefromtheperformanceofmice,asdepictedinthestandardbar-graphrepresentationFigure7.1.AsportrayedinTable7.7fortheinterferencetask,boththeneuralnetwork-andsupportvectormachine-basedclassierswereinferiortoadvancedstatisticalstepwise-forwarddiscriminantanalysismethodsand,moreover,exhibitedsimilarlyidiosyncraticperformancetotheanalysesreportedearlierforthecomprehensivetask199

PAGE 210

battery.Thedecisiontree,however,wastheonlyclassierexaminedwhichwascapa-bleofdistinguishingbetweenTgandTg-KOmiceusingerror-scoredata,utilizingthedelayed-recallmeasure.Inaddition,thedecisiontreewascomparabletostepwise-forwarddiscriminantanalysisfordistinguishingbetweenNTandTg-KOanimals,usingresponse-latencydatawithorwithouterror-scoredata.Theretroactivein-terferencelatencywasshowntobethebestpredictorvariableforgroupmembership,consistentwiththendingsforNTvs.Tg-KOcontrastusingstepwise-forwarddis-criminantanalysisdiscussedearlier.Theperformanceprolesoftheneuralnetworksandsupportvectormachinesareunsatisfactory,relativetothedecisiontrees,andin-terpretationisfurthercomplicatedbymarginalperformancee.g.,oftenonlyslightlyaboverandom-chanceassignmentlevels,aswellasubiquitousnonsignicantsensi-tivityand/orspecicity.Thegoalsofthisstudyweretoexaminetheeectivenessofanovelinterference-basedparadigmforcognitiveassessmentinmice,throughcomparisonwithanes-tablishedcomprehensivebehavioraltaskbattery,andtoutilizeadvancedstatisticalandcomputationaldatamining-basedanalytictechniquesalongwithconventionalANOVA-basedstatisticsapproachesforneurobehavioralresearch.TwogenotypiccomponentsAlzheimer'stransgenicityandGRK5weremanipulated,andthecon-sequentialbehavioralcognitivemanifestationswereevaluatedandanalyzed.AsignicantmaineectfortheAlzheimer'stransgenicgenotypeAPPswwasiden-tiedusingthecomprehensivetaskbattery,throughstandardstatisticaltechniquesforindividualcognitivemeasures,aswellasbyadvancedstatisticalanddatamining-basedclassiers,consistentwithpriorndingse.g.,Leightyetal.,2004;Leightyetal.,2008.Similarly,behavioralmeasuresfromtheinterferenceparadigmsuccessfullydistinguishedbetweenNTandTgmiceusingstandardstatistics,advancedstatisticsanddataminingtechniques.200

PAGE 211

TheeectofGRK5expressionagainstanontransgenicNTbackgroundwasre-vealedwiththecomprehensivetaskbattery,whereinasinglesensorimotormeasureOpenFieldactivitywasidentiedbystandardstatistics,andbothsensorimotorandcognitivemeasureswerereturnedbystepwise-forwarddiscriminantanalysis.Bycon-trast,noneoftheclassiersexaminedwereabletodistinguishbetweenNTandNT-KOindividualsusingbehavioralmeasuresfromtheinterferenceparadigm.Theinu-enceoftheGRK5genotypeagainstanAlzheimer'stransgenicbackgroundAPPswwasindicatedbytwosensorimotormeasuresOpenFieldactivity,ElevatedPlusmazeopen-armentriesthroughstandardstatistics,aswellasbystepwise-forwarddiscrim-inantanalysisanddecisiontreesusingbehavioraldatafromthecomprehensivetaskbattery.Usingmeasuresfromtheinterferenceparadigm,theTgandTg-KOgroupswerelargelyindistinguishablethroughstandardstatisticsanddecisiontree-basedclassiersexceptusingerror-scoredataonly.Additionally,theNTandTg-KOgroupsweredistinguishableusingstandardstatisticsonlyforretroactiveinterfer-enceresponse-latencyintheinterferenceparadigm,aswellasbystepwise-forwarddiscriminantanalysesofboththecomprehensivetaskbatteryrequiringsensorimo-tor,anxiety,andcognitive-basedmeasuresandinterferenceparadigmmeasures.Thisstudyalsodemonstratedthattheeectivenessofgroupclassicationwasde-pendentuponwhetherthecomprehensivetaskbatteryortheinterferenceparadigmwasusedasthesourceforthebehavioralmeasures.BycomparingTables7.4and7.7,forexample,wendthatbehavioralmeasuresfromthecomprehensivetaskbatteryprovidedsuperiordiscriminabilitybyallclassiersbothadvancedstatisticalanddatamining-based,relativetointerferenceparadigm-basedmeasures,fordistin-guishingbetweennontransgenic-controlandAlzheimer'stransgenic-controlanimals,aswellasamongthefourgroupsofmice.Thismaybeattributabletothebroadersamplingfromthebehavioralrepertoiresensorimotor,anxiety,andcognitivecom-201

PAGE 212

ponentsoeredbythecomprehensivebattery,relativetothecognitive-basedinter-ferenceparadigm.Asreportedearlier,successfulgroupwisediscriminabilityisoftenbaseduponacombinationofsensorimotor-based,anxiety-based,and/orcognitive-basedmeasures.Indeed,discriminantanalysis,neuralnetworks,andsupportvectormachineswereabletodistinguishbetweentheNTandNT-KOgroups,aswellasbetweentheTgandTg-KOgroups,usingbehavioralmeasuresfromthecomprehen-sivetaskbattery;however,noneofthesethreetypesofclassierssucceededusinginterferenceparadigmmeasuresinstead.Hereagain,theavailabilityofalargenum-berofcandidatepredictorvariablesincreasedthelikelihoodofdistinguishingbetweengroups.Bycontrast,interferenceparadigmmeasuresspecically,response-latenciesprovidedsuperiordiscriminabilitybetweentheNTandTg-KOgroups,forallclassi-ersexceptsupportvectormachineswhichshowedcomparableperformance.Thechoiceofbehavioralmetricserrorsvs.latenciesmayalsodeterminetheoptimalclassiermodel.FordetectingtheeectofGRK5expressiononanAlzheimer'stransgenicbackgroundi.e.,Tgvs.Tg-KO,forinstance,adecisiontreeprovidedwithbehavioralmeasuresfromthecomprehensivetaskbatterywilloutperformadecisiontreeusingonlyinterferenceparadigm-basedresponse-latencymeasures,butunderperformadecisiontreeusingonlyerror-scoreinterferencemeasures.Finally,neitherdecisiontree-basedclassierusingmeasuresfromthecomprehensivebatteryortheinterferenceparadigmreliablydistinguishedbetweennontransgenicmiceonthebasisofGRK5expressioni.e.,NTvs.NT-KO.TheseresultshaveseveralimportantimplicationsfortransgenicmousemodelsofAlzheimer'sdisease,aswellasbehavioralevaluationandanalyticmethodology:First,theAPPgenotypeundoubtedlyexertsaverypowerfulinuenceonthecogni-tivephenotypeofmicewithinthemixedbackgroundofourcolony.Indeed,asprevi-ouslydiscussed,evenasinglebehavioralmeasuree.g.,RAWMT4errors,delayed-202

PAGE 213

recallerrorsmaybesucienttodistinguishbetweennontransgenicandAlzheimer'stransgenicanimals.Second,thatclassiersexhibitdierentialsensitivitytobehavioralfeatures.Insomecases,error-scoresarebetterdiscriminatorse.g.,Tgvs.NTusingneuralnetworks,whileinothercases,response-latenciesaremoreeectivee.g.,Tgvs.NTusingstepwise-forwarddiscriminantanalysis.Forthisreason,itisextremelyimportantforpractitionerstoexamineandcomparealternativeclassierdesigns,toevaluateapplicabilityforaspecicresearchproblem.Moreover,theimportanceofdiversesamplingacrossthebehavioralrepertoiresensorimotor-based,anxiety-based,andcognitive-basedtasks/measurescannotbeoverstated,tomaximizethechancesofdetectinggroupwisedierencesotherwiseoverlooked.Third,thatadditionalresponsedatamaynotnecessarilyimprovegroupwisedis-criminability,norinterpretabilityofanalyticresults.Forbehavioralmeasuresoftheinterferenceparadigm,theavailabilityofbotherror-scoresandresponse-latenciesdidnotnecessarilyimprovetheperformanceoftrainedclassiers,comparedtousingeitherofthetworesponsemeasuresalone.Inaddition,decisiontreeperformancedeclines,supportvectormachinesimproveslightly,andstepwise-forwarddiscrimi-nantanalysesoftenremainthesamewhenlatencymeasuresareincluded,relativetousingerror-scoresalone.And,Finally,althoughthecomprehensivetaskbatteryprovidesarichersamplingofthemousebehavioralrepertoire,thusgenerallyimprovingdiagnosticeectivenesse.g.,distinguishingbetweentreatmentgroups,themouse-basedinterferenceparadigmdevelopedandrenedinthisStudyrepresentsapowerfulnewinstrumentforcogni-tiveassessmentinAlzheimer'sdiseaseresearchwithtransgenicanimals.203

PAGE 214

CHAPTER8INTERFERENCETASK-BASEDTHERAPEUTICEVALUATIONOFGM-CSF8.1IntroductionGranulocytemacrophagecolony-stimulatingfactorGM-CSFisaninammato-geniccytokinecomponentoftheimmune-inammatorycascade,whichpromotespro-ductionofbothgranulocyteseosinophils,basophils,andneutrophilsandmonocyteswhich,inturn,matureintoGM-CSF-secretingmacrophagesbystemcells.Inhu-mans,GM-CSFisusedtostimulatewhitebloodcellproductionfollowingchemother-apy.However,whenadministeredtoAlzheimer'stransgenicmiceinconjunctionwithIL-4,asanadjuvantinbeta-amyloidimmunotherapy,fewercorticalCongophilicplaquesandreducedplaque-associatedmicrogliosiswerereportedDaSilvaetal.,2006.CoadministrationofGM-CSFandIL-rpromotesanattenuatedTh2responsetobeta-amyloidimmunization,includingantibodieswhich,inturn,reduceplaqueburdenDaSilvaetal.,2006.GM-CSFlevels,measuredinbrainslicecultures,arepositivelycorrelatedwithbeta-amyloidbothA40andA42inAlzheimer'strans-genicmicePateletal.,2005,underscoringtheroleofcytokinesintheneuroin-ammatoryresponsetobeta-amyloid.Successfulattenuationofplaque-associatedstructuralandfunctionalimpairmentmayconstituteatherapeuticroleforGM-CSF,albeitindirect,topreserveorrestorecognitivefunctioninAD.ThepurposeofthisstudywastoinvestigatethecognitiveeectsofGM-CSFadministrationinbothnontransgenicandAlzheimer'stransgenicmiceusinganovel,204

PAGE 215

mouse-basedinterferencelearningparadigm.ThisbehavioralparadigmwasadaptedfromasemanticinterferencetestingprotocolLoewensteinetal.,2004forADdi-agnosisandscreeninginclinicalsettings,bysubstitutingrodent-appropriatetestconditionsRAWMapparatus,visualcuesandspatialmemorydomaininplaceoftheoriginalprotocol'sverbalstimuliandsemanticcomponents,aswellasincludingbotherror-scoreandresponse-latencymeasuresfromeachbehavioraltask.DetailsofthebehavioraltasksintheoriginalprotocolwerepresentedearlierrefertotheInterferenceTestinginHumans:AComparisonofStatisticalandDataMiningMethods"Study,Chapter6.Themouse-basedbehavioralparadigmconsistsoffourinterrelatedtasks,asdescribedinthepreviouschapterTheInterferenceTask:ANovelAssessmentParadigm",Chapter7:Three-trialrecall,proactiveinterference,retroactiveinterference,anddelayed-recall.Behavioralmeasureserror-scoresand/orresponse-latenciesfromoneormoreofthesetaskshavesuccessfullydistinguishedbe-tweennontransgenicandAlzheimer'stransgenicmice,aswellasbetweenAlzheimer'stransgenicanimalswithanadditionalgeneticmodicationGRK5-knockoutandtransgenicshavingtheGRK5-wildtypegenotypeChapter7.TheapplicabilityoftheinterferenceparadigmforevaluatingGM-CSFtherapeuticbenetswasexam-inedusingbothadvancedstatisticaldiscriminantanalysesanddatamining-basedmethodsdecisiontrees,neuralnetworks,supportvectormachines,anddiscussedinlightofstandardstatisticalanalysisofvariance;ANOVAndingsbyArendashetal.unpublisheddata.8.2MaterialsandMethodsThesubjectsconsistedofage-matchedmicegeneratedfromacrossbetweenhet-erozygousAPPswAPPmutationK670NandM671LandheterozygousPS1Tgline6.2mice,fromwhich16Alzheimer'stransgenicAPPswand16nontransgenic205

PAGE 216

NTmicewereselected.Atone-yearofage,these32animalswerecognitivelypre-testedforeightdaysusingtheRAWMworkingmemorytasktoassesscognitiveper-formance.TheAPPswmicewerethenseparatedintotwogroupsbyrandomassign-ment,balancedforcognitiveperformanceandbloodbeta-amyloidlevels:Alzheimer'stransgenic-controlTg;N=8andtransgenic-treatmentTg+GMCSF;N=8.Simi-larly,theNTmicewererandomlyassignedtoeitheroftwogroups,balancedforcog-nitiveperformance:Nontransgenic-controlNT;N=8andnontransgenic-treatmentNT+GMCSF;N=8.Aftertwoweeks,themicebeganreceivingdailysubcuta-neousinjectionsofeitherGMCSFmcgNT+GMCSFandTg+GMCSFgroupsorplainvehicleNTandTggroupsfortendays.Allanimalswerethencognitivelypost-testedforfourdaysusingtheRAWM,followedbytwodayswithouttesting,andconcludedbytheinterferenceparadigmforfourdays.Dailyinjectionswerecontinuedthroughoutthecognitiveevaluationperiod.Behavioralmeasuresofallanimalsfromthenaltwo-dayblockofthefour-dayRAWMpost-testi.e.,botherrorsandlatenciesforworkingmemory,trialsT4andT5,aswellasfromthenaltwo-dayblockofthefour-dayinterferenceparadigmi.e.,botherrorsandlatenciesforthree-trialrecall,proactiveinterference,retroactiveinterference,anddelayed-recall,werecompiledforsubsequentanalyses.StandardstatisticalmethodsANOVA,F-testwereusedtoidentifygroupdierencesinthenal-blockpost-testRAWMdataset,aswellasthenal-blockinterferencedataset.Inaddition,performancemeasuresfromthenal-blockoftheinterferenceparadigmwereusedtogenerateadditionaldatasetsforfurtheranalysis,asdescribedintheMouseInterferenceParadigminSection7.2oftheTheInterferenceTask:ANovelAssessmentParadigm"StudyChapter7.Theeightcognitivemeasures,consistingofmeanerror-scoreandresponse-latencyfromeachofthefourtasksoftheinterfer-enceparadigm,weregroupedintothreeseparatedatasetserrors-only,latencies-only,206

PAGE 217

botherrorsandlatenciesforanalysisusingbothadvancedstatisticalcorrelationanalysis,factoranalysis,anddiscriminantanalysisanddatamining-basedmethodsdecisiontrees,neuralnetworks,andsupportvectormachines.Completespeci-cationsofthecomputinghardwareandsoftwareutilizedintheanalyses,includingprogramparametersettings,areprovidedintheGeneralAnalyticProtocolofSection4.2intheCaeineAdministrationinNontransgenicMice"StudyChapter4.AnimalcareandusewasinaccordancewiththeGuideandUseofLaboratoryAn-imals,NationalResearchCouncil,1996,inaprogramandfacilitiesfullyaccreditedbytheAssociationforAssessmentandAccreditationofLaboratoryAnimalCare,International,underaprotocolapprovedbytheUniversityofSouthFloridaInsti-tutionalAnimalCareandUseCommitteeNo.3183,HuntingtonPotter,Ph.D.,PrincipalInvestigator.207

PAGE 218

8.3ResultsofStatisticalAnalyses8.3.1StandardStatisticalAnalysisFigure8.1nextpageshowsthecognitiveperformanceerror-scoresofallgroupsinthepost-testRAWMworkingmemorytrialsT4andT5forbothtwo-dayblocks.Groupwisemeansandstandard-errorareindicatedinthegure.Thedouble-asteriskdenotessignicantdierencebetweentheTggroupandtheotherthreegroupsofanimalsintheRAWMworkingmemorytrialT5measurep<.05.Thedaggersym-bolyindicatessignicantcontrastp<.05withNT+GMCSFandTg+GMCSF.MeasuredtendaysintotheGMCSFtreatmentperiod,Tg-controlmiceexhibitedim-pairmentinRAWMworkingmemorytrialsT4andT5,comparedwithNTanimals.CognitiveimpairmentofTg-controlmicewasevidentwithintwo-dayblocksFigure8.1,upper,aswellasacrossallfourdaysofRAWMtestingFigure8.1,lower.Alzheimer'stransgenicanimalsreceivingGM-CSFTg+GMCSF,bycontrast,per-formedcomparablyor,indeed,superiortoNTonRAWMworkingmemorytrialsT4andT5,bothwithinblocksandoverall.Hence,GM-CSFwasshowntore-verseworkingmemoryimpairmentasevaluatedbytheRAWMtaskinAlzheimer'stransgenicmice.208

PAGE 219

Figure8.1.Groupwisecomparisonoferror-scoresinRAWMpost-testtrialsT4andT5,byblocks,intheGMCSFstudy209

PAGE 220

Figure8.2nextpageportraysgroupwisecognitiveperformanceonallmeasuresofthenal-blockoftheinterferenceparadigm,indicatingthemeanandstandarder-rorassociatedwitheachmeasureforallgroups.StandardstatisticaltestsANOVA,F-testrevealasignicantdierencep<.05betweentheTg-controlgroupandtheotherthreegroups,denotedbyadouble-asterisk,aswellasasignicantdierencep<.05betweentheTg-controlandNT+GMCSFgroupsindicatedbyasingleasterisk,.Withrespecttobotherror-scoreandresponse-latencymea-sures,Tg-controlmiceshowedsignicantimpairmentinboththethree-trialrecallanddelayed-recalltasksoftheinterferenceparadigm,relativetoGMCSF-treatedAlzheimer'stransgenicmiceTg+GMCSFandbothnontransgenicgroupsNT,NT+GMCSF.Indeed,GMCSFtreatmentinTgmicedramaticallyimprovednor-malizedtheircognitiveperformanceinboththethree-trialrecallanddelayed-recalltaskstolevelscomparabletoNTanimals.Additionally,theTg-controlgroupwassig-nicantlyimpairedcomparedwiththeNT+GMCSFgroup,withrespecttoretroac-tiveinterferencebotherrorsandlatencies.Finally,GMCSF-treatednontransgenicanimalsNT+GMCSFperformedbetteri.e.,bothdecreasederrorsandlatenciesintheproactiveinterferenceandretroactiveinterferencetasks,relativetountreatednontransgenicsNT,althoughthesedierencesdidnotmeetthecriterionforsignif-icancebothp'sbetween.05and.20.210

PAGE 221

Figure8.2.GroupwisecontrastsforallinterferenceparadigmmeasuresintheGMCSFstudy211

PAGE 222

8.3.2CorrelationAnalysisTable8.1.CorrelationsbetweenbehavioralmeasuresoftheinterferenceparadigmintheGMCSFstudy TT-L .93 .000 PI-E PI-L .93 .000 RI-E .62 .56 .73 .69 .000 .001 .000 .000 RI-L .55 .58 .67 .74 .92 .001 .000 .000 .000 .000 DR-E .77 .64 .47 .36 .74 .64 .000 .000 .007 .044 .000 .000 DR-L .79 .77 .48 .44 .74 .74 .93 .000 .000 .005 .011 .000 .000 .000 TT-E TT-L PI-E PI-L RI-E RI-L DR-E Table8.1showssignicantp<.05pairwisecorrelationsobservedbetweenbe-havioralmeasuresforallgroups.Thecorrelationcoecientr-valueandcorre-spondingsignicancep-valueareindicatedatthetopandbottom,respectively,withinmarkedcellsofthetable.Thefollowingabbreviationsareusedtorepresentthefourtasksofthemouse-basedinterferenceparadigm:Three-trialrecall,errorsandlatencyTT-E,TT-L;proactiveinterference,errorsandlatencyPI-E,PI-L;retroactiveinterference,errorsandlatencyRI-E,RI-L;and,delayed-recall,errorsandlatencyDR-E,DR-L.Signicantpositivecorrelationswereobservedbetweentheerror-scoremeasuresofthethree-trialrecallandbothretroactiveinterferenceanddelayed-recalltasks.Asimilarpatternofassociationwasfoundamongthecorrespondinglatencymeasuresofthesethreetasks.Thesendingsindicateclosecorrespondencebetweentasksin-volvingPoolAofthebehavioralparadigm.Theerror-scoreandresponse-latency212

PAGE 223

measuresofeachofthefourtasksweresignicantlycorrelated,aswell,underscoringthecomparabilityofbothindicesofcognitiveperformanceforthetaskscompris-ingthemouse-basedinterferenceparadigm.Additionally,bothmeasuresfromthethree-trialrecalltaskweresignicantlyandpositivelyintercorrelatedwithbothmea-suresofretroactiveinterference,suggestingsuccessfulacquisitionandretentionoftheinitiallearningcomponentPoolA,despitethebriefinterveningdistractortaskY-mazeexposureandinterferencetaskPoolBexposure.Similarly,completein-tercorrelationbetweenallmeasuresoftheretroactiveinterferenceanddelayed-recalltaskssuggestslong-termstabilityoflearning.Thisinterpretationissupportedbysignicantpairwiseintercorrelationamongallmeasuresofthethree-trialrecallanddelayed-recalltasks.Insharpcontrast,bothbehavioralmeasuresfromtheproactiveinterferencetaskdidnotcorrelatewithbothmeasuresfromtheinitialthree-trialre-call.ThisindicatesindependentperformancebetweenPoolAthelearningconditionforthree-trialrecallandPoolBthelearningconditionforproactiveinterference.However,bothbehavioralmeasuresfromproactiveinterferenceweresignicantlycor-relatedwithallmeasuresfromboththeretroactiveinterferenceanddelayed-recalltasks.Foranimalsexhibitinglowerrors/latenciesacrossallthreeofthesetasks,thiswouldindicateexibilityoflearningacrossdierentproblemcontextsPoolAvs.PoolB{thatis,theanimalswereabletorecalltheoriginallearningPoolA,despiteinterference/distractorelementsandatimedelay,aswellastorapidlyswitchbetweenthetwolearningconditions.ThecorrelationmatrixobtainedinthepresentstudydiersfromthatobtainedinthepreviousstudyChapter#7,inthatsignicantpairwisecorrelationswerefoundinthepresentstudybetweenmeasuresfromtheProactiveInterferencetaskandmea-suresfromboththeRetroactiveInterferenceandDelayed-RecallTasks.Thisimpor-213

PAGE 224

tantdierencehasimplicationsfortheresultingfactorstructure,tobediscussedbelow.8.3.3FactorAnalysisTable8.2.Varimax-rotatedfactoranalysisofinterferenceparadigmmeasuresintheGMCSFstudy Factor Measure I II Three-TrialRecallErrors 0.941 Three-TrialRecallLatency 0.897 Delayed-RecallLatency 0.870 Delayed-RecallErrors 0.837 RetroactiveInterferenceErrors 0.736 RetroactiveInterferenceLatency 0.749 ProactiveInterferenceLatency 0.959 ProactiveInterferenceErrors 0.932 Variance 47.51% 39.70% Table8.2depictsavarimax-rotatedprincipalcomponentanalysisoftheeightbe-havioralmeasuresofthemouse-basedinterferencetask.Signicantabsolutevaluegreaterthan0.700componentloadingsareshownforthetwofactorsreturned.Thecalculatedeigenvalueresultsi.e.,>1criterionwereconsistentwithscreeplotCat-tell,1966identicationoftwosignicantfactors.Measuresfromthethree-trialrecallanddelayed-recalltaskscomprisetheprimaryfactor,accountingforapproxi-mately48%ofoverallvariance.ThesetwotasksinvolvespatiallearningwithintheinitialcontextPoolAandlong-termreferencememoryforthissameinformation.Hence,thisfactormayrepresentsuccessfulorunsuccessfulacquisitionandconsoli-dationoftheinitialspatiallearningproblemPoolA.Bycontrast,thesecondfactorrepresentingapproximately40%ofvarianceconsistsofallmeasuresfromtheinter-ferencecomponentsbothproactiveandretroactiveoftheparadigm.Thissecond214

PAGE 225

factormayreectthesubjects'facilityofcontext-switchingPoolAvs.PoolBortheabilitytorapidlyadapttochanginglearningconditions.Asnotedpreviously,dierencesincorrelationpatternsamongbehavioralmea-sures,observedbetweenthisstudyandtheprecedingoneChapter#7,arereectedinthecorrespondingfactorstructures.Incontrasttotheprecedingstudy,FactorIIofthepresentstudyalsoincludesbothbehavioralmeasuresfromtheRetroactiveInterferencetask,whichisconsistentwiththeobservedpatternofcorrelationTable8.1.Interestingly,whilebothmeasuresfromtheDelayed-RecalltaskloadedonaseparatefactorfrombothmeasuresoftheProactiveInterferencetask,inbothstud-ies,signicantpairwisecorrelationsbetweenmeasuresoftheProactiveInterferenceandDelayed-Recalltaskswereonlyfoundinthepresentstudy.Thecauseofthisdis-crepancyisunclear,althoughtreatment-biaseectsand/ordisproportionatesamplesizesunliketheprecedingstudy,alltreatmentgroupsofthepresentstudyconsistedofidenticalnumbersofmicearepotentialsources.215

PAGE 226

Table8.3.ClassierperformancecomparisonusinginterferenceparadigmmeasuresintheGMCSFstudy Evaluation DiscriminantAnalysis Decision Neural Groups Dataset Criterion Complete Step-Fwd Tree Network SVM Accuracy 88% 94% 88% 75% 88% Errors Sensitivity 88% 100% 100% 75% 88% Specicity 88% 88% 75% 75% 88% Tg Accuracy 81% 81% 56% 81% 81% vs. Latency Sensitivity 88% 75% NS 75% 75% NT Specicity 75% 88% 75% 88% 88% Accuracy 88% 94% 88% 69% 75% Err+Lat Sensitivity 100% 100% 100% 63% 63% Specicity 75% 88% 75% 75% 88% Accuracy NS NS 56% 63% 63% Errors Sensitivity NS NS 63% 63% 63% Specicity NS NS NS 63% 63% NT Accuracy NS 69% NS 56% 69% vs. Latency Sensitivity NS 63% NS 63% 63% NT+ Specicity NS 75% NS NS 75% GMCSF Accuracy NS 69% NS 56% 56% Err+Lat Sensitivity NS 63% NS NS 63% Specicity NS 75% NS 63% NS Accuracy 81% 94% 81% 100% 75% Errors Sensitivity 100% 100% 88% 100% 100% Specicity 63% 88% 75% 100% NS Tg Accuracy 94% 94% NS 88% 75% vs. Latency Sensitivity 88% 88% NS 88% 75% Tg+ Specicity 100% 100% NS 88% 75% GMCSF Accuracy 88% 94% 81% 94% 82% Err+Lat Sensitivity 88% 88% 88% 100% 100% Specicity 88% 100% 75% 88% 63% Errors Accuracy 47% 53% 50% 48% NS Allfour Latency Accuracy 41% 38% NS NS 28% Err+Lat Accuracy 47% 59% 41% 41% NS 216

PAGE 227

8.3.4DiscriminantAnalysisAsummarycomparisonofgroupwisediscriminabilityisshowninTable8.3,forreference.Thetablereportstheclassierperformanceevaluationi.e.,accu-racy,sensitivity,andspecicityforadvancedstatisticaltechniquesdirect-entryandstepwise-forwarddiscriminantanalysesinthefourthandfthcolumns.NontransgenicNTvs.TransgenicTgGroupsError-ScoresOnly.AsreportedinTable8.3,direct-entrycompletediscrimi-nantanalysisreturnedexcellentdiscriminabilityWilks'lambda=.261,p=.0031betweentransgenic-andnontransgenic-controlanimalsaccuracy=88%,sensitiv-ity=88%,specicity=88%onthebasisofthefourpredictorvariables.Thestepwise-forwardapproachgeneratedamoreparsimoniousmodel,consistingofonlytwopredictorvariablesthree-trialrecallandretroactiveinterference,withsigni-cantdiscriminabilityWilks'lambda=.271,p=.0002.Thesimplertwo-variablemodeldemonstratedoutstanding4%accuracy,100%sensitivity,and88%speci-cityfortransgenicityi.e.,alltransgenicanimalswerecorrectlyidentied.Response-LatenciesOnly.Usingonlylatencydatafromthefourtasks,direct-entrydiscriminantanalysissuccessfullydistinguishedbetweenthetwogroupsWilks'slambda=.383,p=.0225,with81%accuracysensitivity=88%,specicity=75%,asshowninTable8.3.Thestepwise-forwardapproachalsodemonstratedsignicantdiscriminabilityWilks'lambda=.452,p=..0058betweentransgenic-andnontransgenic-controlanimalsaccuracy=81%,sensitivity=75%,specicity=88%,althoughtheoverallaccuracywascomparabletothedirect-entryclassierresult.Thestepwise-forwardmodelretainedonlythethree-trialrecallandproactiveinterferencelatencymeasures.217

PAGE 228

ErrorsandLatencies.Adirect-entrydiscriminantanalysisshowedsignicantdiscriminabilitybetweengroupsusingbotherror-scoresandreponse-latenciesWilks'lambda=.079,p=.0031betweentransgenic-andnontransgenic-controlanimalsaccuracy=88%,sensitivity=100%,specicity=75%usingalleightmeasuresaspredictorvariables.Bycontrast,thestepwise-forwardapproachretainedonlythreepredictorvariablesthree-trialrecallerrors,retroactiveinterferenceerrors,andthree-trialrecalllatency,anddisplayedsignicantdiscriminabilityWilks'lambda=.143,p=.0000betweenthetwogroups,asreportedinTable8.3.Thesimplerthree-variablemodeldemonstratedsuperior4%accuracy,100%sensitivity,and88%specicityfortransgenicity.Summary:NontransgenicandAlzheimer'stransgenicanimalswerereliablydis-tinguishedusingeither/botherror-scoresandresponse-latenciesobtainedfrombe-havioraltasksintheinterferenceparadigm.Theerrormeasuresfromthethree-trialrecallandretroactiveinterferencetaskswereemphasizedbythediscriminantanaly-ses,regardlessoftheavailabilityofresponse-latencymetrics.NontransgenicNTvs.Nontransgenic-TreatmentNT+GMCSFGroupsError-ScoresOnly.Neitherdirect-entrycompletenorstepwise-forwarddis-criminantanalysissuccessfullydistinguishedbetweenGMCSF-treatedanduntreatednontransgenicmiceusingonlyerror-scoremeasuresfromtheinterferenceparadigm.Noneofthecandidatepredictorvariablesmetthestatisticalcriterionalpha-to-enterforinclusioninthemodel.Response-LatenciesOnly.Althoughsignicantdiscriminabilitywasnotdemon-stratedusingdirect-entrydiscriminantanalysiswithresponse-latencydata,asin-dicatedinTable8.3,thestepwise-forwardapproachreturnedsignicantdiscrim-inabilityWilks'lambda=.622,p=.0457betweenthetwogroupsaccuracy=69%,sensitivity=63%,specicity=75%,withrespecttoGMCSFtreatment.The218

PAGE 229

stepwise-forwardmodelretainedboththethree-trialrecallandproactiveinterferencemeasures.ErrorsandLatencies.Direct-entrydiscriminantanalysiswasunabletodistin-guishbetweenthetwogroupsusingbotherror-scoresandreponse-latencies.How-ever,thestepwise-forwardapproachreturnedsignicantdiscriminabilityWilks'lambda=.622,p=.0457betweenthetwogroupsusingatwo-variablemodelretainingthelatencymeasuresfromthethree-trialrecallandproactiveinterferencetasks,asshowninTable8.3.Thetwo-variablemodelshowed69%accuracy,63%sensitivity,and75%specicity,withrespecttotreatment.Summary:ThedicultyindistinguishingbetweenuntreatedandGMCSF-treatednontransgenicanimalswasunderscoredbytheinabilityofdirect-entrydiscriminantanalysistoidentifyindividualsbygroupusingbehavioralerrorand/orlatencymea-sures.Althoughstepwise-forwarddiscriminantanalysesconsistentlyidentiedthethree-trialrecallandproactiveinterferenceerror-scoremeasuresasthebestpredic-torvariablesfordistinguishingbetweenGMCSF-treatedanduntreatednontrans-genicmice,therelativelypoordiscriminabilitybetweenthesetwogroupsusingeither/botherrorsandlatenciessuggeststhatGMCSFtreatmentdoesnotaectnormalmice.TransgenicTgvs.Transgenic-TreatmentTg+GMCSFGroupsError-ScoresOnly.Direct-entrycompletediscriminantanalysissuccessfullydistinguishedWilks'lambda=.276,p=.0042betweenthetwogroupsaccuracy=81%,sensitivity=100%,specicity=63%,asindicatedinTable8.3.Themodelgeneratedbythestepwise-forwardapproachincludedthethree-trialrecallandproactiveinterferenceerror-scoresaspredictorvariables.Thistwo-variablemodelalsoshowedsignicantdiscriminabilitybetweenthetwogroupsWilks'lambda=219

PAGE 230

.335,p=.0008,withhigh%accuracy,100%sensitivity,and88%specicity,withrespecttotreatment.Response-LatenciesOnly.Signicantdiscriminabilitywasdemonstratedusingdirect-entrydiscriminantanalysiswithresponse-latencydataWilks'lambda=.236,p=.0018,asreportedinTable8.3.Theresultingclassierachieved94%overallaccuracy,88%sensitivity,and100%specicity,withrespecttotreatment.Thestepwise-forwardapproachalsoreturnedsignicantdiscriminabilityWilks'lambda=.273,p=.0002,andusedonlytheproactiveinterferenceanddelayed-recalllatencymeasuresaspredictorvariables.Thisparsimoniousmodeldemonstrated94%accuracy,88%sensitivity,and100%specicity,withrespecttotreatment.Alltransgenic-controlanimalswerecorrectlyidentiedbybothclassiers.ErrorsandLatencies.Direct-entrydiscriminantanalysisusingbotherror-scoresandresponse-latenciessuccessfullydistinguishedbetweenthetwogroupsWilks'lambda=.066,p=.0017betweenthetwogroupsaccuracy=88%,sensitivity=88%,specicity=88%,asdepictedinTable8.3.Threevariableswereretainedbythestepwise-forwardapproachproactiveinterferenceerrorsandlatency,delayed-recalllatency.Thethree-variablemodelalsodistinguishedbetweenGMCSF-treatedanduntreatedAlzheimer'stransgenicmiceWilks'lambda=.191,p=.0001,withexcellent94%accuracy,88%sensitivity,and100%specicity,withrespecttotreat-ment.Summary:Excellent%discriminabilitywasachievedusingeither/botherror-scoresandresponse-latencies,consistentwithsignicantdierencesincognitiveper-formancebetweenthetwogroupsofmice.Error-scoredataalone,however,wasmoreeectiveforcorrectlydetectingtheGMCSFtreatmenteectinAlzheimer'stransgenicmicei.e.,greatersensitivity.220

PAGE 231

AllFourGroupsError-ScoresOnly.Signicantdiscriminabilityamongthefourgroupswasre-turnedbydirect-entrycompletediscriminantanalysisWilks'lambda=.274,p=.0005,with47%overallaccuracy,asindicatedinTable8.3.TheGMCSF-treatednontransgenicanimalswerethemost-frequentlymisclassiedindividuals,withonly25%correctlyidentiedrandom-chancelevelperformance.Bycontrast,thestepwise-forwardapproachdemonstratedhigherdiscriminabilityWilks'lambda=.383,p=.0002andaccuracy3%usingonlytwoerrormeasures:Three-trialrecallandproactiveinterference.Response-LatenciesOnly.Direct-entrydiscriminantanalysissuccessfullydistin-guishedamongthefourgroupsusingonlyreponse-latencydataWilks'slambda=.391,p=.0143,andexhibited41%overallaccuracy.Interestingly,althoughthestepwise-forwardapproachalsoreturnedsignicantdiscriminabilityamongthegroupsWilks'slambda=.473,p=.0019,itsoverallaccuracywasonly38%.Thestepwise-forwardmodelselectedboththethree-trialrecallandproactiveinterferencelatencymeasures.TheGMCSF-treatedAlzheimer'stransgenicmicewerethemostlikelytobemisclassied;only25%oftheseanimalswerecorrectlyidentied,asifnormalized"bycontrastingwiththeNTandNT+GMCSFanimals.ErrorsandLatencies.Adirect-entrydiscriminantanalysisreturnedsignicantdiscriminabilityWilks'slambda=.119,p=.0007amongthefourgroupsaccu-racy=47%,asshowninTable8.3,usingtheeightbehavioralmeasuresaspredictorvariables.Thestepwise-forwardapproachretainedonlythreemeasuresthree-trialrecallerrorsandlatency,proactiveinterferencelatency,andshowedsignicantdis-criminabilityWilks'slambda=.287,p=.0001andrelativelyhigh%overallaccuracyamongthegroups.Only30%oftheGMCSF-treatednontransgenicmicewerecorrectlyidentied;theseanimalswerethemostlikelytobemisclassied,reec-221

PAGE 232

tiveofnormalized"performancetheNT,NT+GMCSF,andTg+GMCSFanimalsperformedcomparably.Summary:Variance-optimizedstepwise-forwarddiscriminantanalysis-basedclas-siersusingbotherror-scoreandresponse-latencymeasuresexhibitbettergroupwisediscriminabilitythanclassiersbasedoneithererrorsorlatenciesalone.Moreover,therelativelypooroverallaccuracymaybeattributedtothedicultyindistin-guishingamongNT,NT+GMCSF,andTg+GMCSFgroups,duetosimilaritiesincognitiveperformance.8.4ResultsofDataMiningAnalysesTable8.3comparesgroupwisediscriminabilityforeachclassierinthepresentstudy.Performanceevaluationisreportedi.e.,accuracy,sensitivity,andspecicitywhensignicantexceedingrandom-chancelevelassignmentfordatamining-basedmethodsdecisiontrees,neuralnetworks,andsupportvectormachinesinthesixththrougheighthcolumns.8.4.1DecisionTreeAnalysisNontransgenicNTvs.TransgenicTgGroupsError-ScoresOnly.Whenprovidedwithonlyerror-scorebehavioraldatafromalltasksoftheinterferenceparadigm,thedecisiontree-basedclassierselectedasingleattribute,thethree-trialrecallmeasure,asthecriterionforsplittingthedatasetintotwogroups.AsshowninTable8.3,theresultingtreecorrectlyidentied88%Kappa=0.75ofindividuals,with100%sensitivityi.e.,allTgmicewerecorrectlyidentiedand75%specicityi.e.,allbuttwoNTanimalswerecorrectlyidentied.Response-LatenciesOnly.Thedecisiontreeidentiedthreebehaviorallatencymeasures{delayed-recall,proactiveinterference,andthree-trialrecall{forsplitting222

PAGE 233

thedatasetintotwogroups.Theclassierdisplayedaverymodest56%accuracyKappa=0.13,with38%sensitivitynotsignicantand75%specicity.Hence,thisclassierwasmoreappropriateforcorrectlyidentifyingnontransgenicindividu-als.ErrorsandLatencies.Whenbotherror-scoresandresponse-latencieswereusedtoconstructthedecisiontree-basedclassier,onlythethree-trialrecallerrormeasurewasselectedashavingsucientinformation-biastosplitthedatasetintothetwogroups.Theclassiercorrectlyidentied88%ofindividualsKappa=0.75,with100%sensitivityand75%specicity,asindicatedinTable8.3.Thus,theinclusionoflatencymeasureswitherrormeasuresdidnotchangetheoverallclassicationaccuracy.Summary:Error-scores,eitheraloneortogetherwithresponse-latencymeasures,provideexcellentdiscriminabilitybetweennontransgenicandAlzheimer'stransgenicmice,usingtheinterferenceparadigm.ThisresultunderscoresthecognitivedecitspresentinTgmice,relativetonontransgenics,basedonconventionalprotocolsforevaluatione.g.,RAWM.NontransgenicNTvs.Nontransgenic-TreatmentNT+GMCSFGroupsError-ScoresOnly.Thedecisiontreeselectedasinglebehavioralmeasure,retroac-tiveinterferenceerrors,asthebestattributeforsplittingthedatasetintotwogroups,usingonlybehavioralerror-scoredata.AsshowninTable8.3,theclassierdemon-strated56%accuracyKappa=0.13,63%sensitivity,and50%specicitynon-signicant,withrespecttotreatment.One-halfoftheuntreatednontransgenicanimalsweremisclassiedasGMCSF-treated."Response-LatenciesOnly.Noneoftheresponse-latencymeasuresprovidedsu-cientinformationforconstructingadecisiontree-basedclassiertodistinguishbe-tweenGMCSF-treatedanduntreatednontransgenicmice.223

PAGE 234

ErrorsandLatencies.Whenbotherror-scoresandresponse-latencieswerepro-vided,thedecisiontreewasunabletoidentifyanyattributeswhichreliablydistin-guishbetweenthetwogroups.Summary:Althoughtheretroactiveinterferenceerror-scoremeasureprovidedmodestdiscriminabilitybetweenthesetwogroups,theseresultsgenerallysuggestthatGMCSF-treatedanduntreatednontransgenicmicedonotdiersignicantly,withrespecttoperformanceintheinterferenceparadigm.TransgenicTgvs.Transgenic-TreatmentTg+GMCSFGroupsError-ScoresOnly.Thedecisiontreeidentiedasingleattribute,three-trialrecallerrors,forsplittingthedatasetintotwogroups,correspondingtoGMCSF-treatedanduntreated-controlAlzheimer'stransgenicmice.AsreportedinTable8.3,theclassieraccuratelyidentied81%oftheindividualsKappa=0.63,with88%sensitivityand75%specicity.Response-LatenciesOnly.Noneoftheresponse-latencymeasuresprovidedsu-cientinformation-biastoaccuratelydistinguishbetweenthetwogroups.ErrorsandLatencies.Asingleattribute,three-trialrecallerror,wasidentiedasprovidingsucientinformationtodistinguishbetweenthetwogroups,whenallbehavioralmeasuresbotherrorsandlatencieswereavailabletotheclassier.Theresultingtreeexhibited81%accuracyKappa=0.63,with88%sensitivityand75%specicity.ThisisidenticaltotheresultobtainedforError-ScoresOnly.Summary:Error-scoredata,eitheraloneortogetherwithresponse-latencymea-sures,accurately%distinguishesbetweenAlzheimer'stransgenicmicewhichreceivedGM-CSFtreatmentandTganimalswhichdidnot.ThisresultsuggestsaGM-CSFtreatmenteect,withrespecttocognitiveperformanceintheinterfer-enceparadigm,andaprimacyoferror-basedperformancecriteriaoverlatency-basedmeasuresfordetectingGM-CSFtreatmentecacy.224

PAGE 235

AllFourGroupsError-ScoresOnly.Theerror-scoresfromtwobehavioraltasks,three-trialrecallandretroactiveinterference,wereselectedbythedecisiontreeasprovidingoptimaldiscriminabiliityamongthefourgroupsofanimals.Theoverallaccuracy,asreportedinTable8.3,was50%Kappa=0.33.Nontransgenic-controlindividualsweretheleast-likelytobecorrectlyidentied,withonly25%ofNTmicerecognized.Response-LatenciesOnly.Thedecisiontree-basedclassierwasunabletosuc-cessfullydistinguishamongthefourgroupsusingonlyresponse-latencydata.Indeed,noneofthenontransgenic-controlanimalswerecorrectlyidentied.ErrorsandLatencies.Theinclusionofbotherror-scoreandresponse-latencymeasuresforconstructingdecisiontreesresultedin41%accuracyKappa=0.21usingfourattributes:Three-trialrecallerrorsandlatency,retroactiveinterferenceer-rors,anddelayed-recalllatency.However,only13%ofnontransgenic-controlanimalswerecorrectlyidentiedbythisclassier.Summary:Althoughtheoveralldiscriminabilityamongthefourgroupswasrela-tivelypoor,thismaybeduetodicultywithdistinguishingamongNT,NT+GMCSF,andTg+GMCSFanimals.However,asreportedinTable8.3,error-scoresaloneprovidedthebestfour-waydiscriminability0%amongthedecisiontree-basedclassiers.8.4.2NeuralNetworkAnalysisNontransgenicNTvs.TransgenicTgGroupsError-ScoresOnly.Whenprovidedwitherror-scoremeasuresfromalltasksoftheinterferenceparadigm,theneuralnetwork-basedclassiercorrectlyassigned75%oftheindividualsintotheirrespectivegroupsKappa=0.50,withoptimalperfor-manceobtainedwiththreecomputingelementsinthehiddenlayerofthenetwork.225

PAGE 236

Thesensitivitywas75%andspecicitywas75%forthisclassier,asindicatedinTable8.3.Indeed,thesameproportionofanimalsfromeachgroupweremisclassied.Response-LatenciesOnly.Usingonlybehavioralresponse-latencydata,theneu-ralnetworkclassierexhibited81%accuracyKappa=0.63,with75%sensitivityand88%specicity.Aslightlygreaterproportionofnontransgenicanimalswerecorrectlyidentied,relativetotheproportionoftransgenics.ErrorsandLatencies.Only69%ofindividualsKappa=0.38werecorrectlyidentiedbyaneuralnetwork-basedclassiertrainedwithbotherror-scoresandresponse-latencydata.AsreportedinTable8.3,thesensitivityoftheclassierwas63%andthespecicitywas75%.Summary:Theoptimalneuralnetwork-basedclassierfordistinguishingbetweennontransgenicandAlzheimer'stransgenicmiceusedonlyresponse-latencydata,andshowedaslightperformancebiasfavoringnontransgenicanimals,withrespecttocorrectidenticationbygenotype.NontransgenicNTvs.Nontransgenic-TreatmentNT+GMCSFGroupsError-ScoresOnly.Usingonlybehavioralerror-scoredata,theneuralnetwork-basedclassiercorrectlyassignedonly63%oftheindividualsintotheirrespectivegroupsKappa=0.25,with63%sensitivityand63%specicity.Hence,approxi-matelytwo-thirdsoftheanimalsineachgroupwerecorrectlyidentied.Response-LatenciesOnly.Only56%accuracyKappa=0.13wasachievedbyneuralnetworkclassierstrainedusingonlybehavioralresponse-latencydatafromtheinterferenceparadigm.Theobservedsensitivityoftheclassierwas63%,anditsspecicitywas50%.Hence,random-chancelevelclassicationwasobservedfortheuntreatednontransgenicanimals.ErrorsandLatencies.Theneuralnetwork-basedclassiercorrectlyassignedonly56%oftheindividualsKappa=0.13intotheirrespectivegroups,with50%sensi-226

PAGE 237

tivityand63%specicity,asshowninTable8.3.Optimalperformancewasobtainedwithaneuralnetworkarchitecturecontainingfourcomputingelementsinthehiddenlayer.Summary:TherewasdicultydistinguishingbetweenGMCSF-treatedandun-treatednontransgenicmice,althoughmodestdiscriminabilitywasobservedinneuralnetwork-basedclassierstrainedusingbehavioralerror-scoredatafromtheinterfer-enceparadigm.TransgenicTgvs.Transgenic-TreatmentTg+GMCSFGroupsError-ScoresOnly.Aneuralnetwork-basedclassiertrainedusingonlybehav-ioralerror-scorescorrectlyassigned100%oftheindividualstotheirrespectivegroupsKappa=1.00,withbothsensitivityandspecicityof100%,asshowninTable8.3.Optimalperformanceofthisclassierwasachievedwiththreecomputingelementsinthehiddenlayerofthenetwork.Response-LatenciesOnly.Usingonlybehavioralresponse-latencydatatotrainaneuralnetwork-basedclassierresultedinalower8%accuracyKappa=0.75,with88%sensitivityand88%specicity.Thisneuralnetworkrequiredfourhiddenlayerelementsforoptimalperformance.ErrorsandLatencies.Theneuralnetwork-basedclassiercorrectlyassigned94%oftheindividualsintotheirrespectivegroupsKappa=0.88,withoptimalper-formanceobtainedwithfourcomputingelementsinthehiddenlayerofthenet-work.Thesensitivitywas100%andspecicitywas88%.Hence,allGMCSF-treatedAlzheimer'stransgenicanimalswerecorrectlyidentied.Summary:Neuralnetwork-basedclassierstrainedwithbehavioraldatafromtheinterferenceparadigmareextremelyeectivefordistinguishingbetweenGMCSF-treatedanduntreatedAlzheimer'stransgenicmice.Additionally,theeectofnet-worktopologynumberofcomputingelementswithinthehiddenlayeronclassier227

PAGE 238

performancesuggestsdierentunderlyingcomplexityassociatedwithlatencydata,relativetoerrordata.AllFourGroupsError-ScoresOnly.AsindicatedinTable8.3,theneuralnetwork-basedclassierdemonstrated48%overallaccuracyKappa=0.29inassigninganimalstotheirre-spectivegroupsonthebasisofbehavioralerror-scores.Anetworkarchitecturewithfourcomputingelementsinthehiddenlayerwasnecessaryforoptimalperformance.Thepoorestdiscriminabilitywasobservedforbothnontransgenicgroups,whereinonly38%ofindividualsfromeachgroupNT,NT+GMCSFwerecorrectlyidenti-ed,underscoringthesimilarityofthesegroupswithrespecttocognitiveperformanceintheinterferenceparadigm.Response-LatenciesOnly.Theneuralnetwork-basedclassierwasunabletodis-tinguishamongthefourgroupsofmiceusingonlybehavioralresponse-latencydata.Indeed,noneofthenontransgenic-controlNTanimalswerecorrectlyidentied.ErrorsandLatencies.Usingbotherror-scoresandresponse-latencies,41%oftheanimalswerecorrectlyassignedtotheirrespectivegroupsKappa=0.21.TheGMCSF-treatedAlzheimer'stransgenicmicewerethemostlikelyindividualstobecorrectlyidentied%,whileonly38%ofindividualsineachoftheotherthreegroupswerecorrectlyassignedtotheirrespectivegroup.Fourcomputingelementswerepresentinthehiddenlayeroftheoptimalnetwork.Summary:Theoptimalneuralnetwork-basedclassierfordistinguishingamongthefourgroupsusedonlybehavioralerror-scoredata,providedpooroveralldis-criminability8%,andshowedaperformancebiasfavoringthetransgenicbothGMCSF-treatedanduntreatedsubjects.Theinclusionofresponse-latencydatacompromisedoveralldiscriminabilityamonggroups.Additionally,theoptimalnet-worktopologiesrequiredfourcomputingelementsinthehiddenlayer,regardlessof228

PAGE 239

howmanycandidatepredictorvariablesbehavioralmeasuresweresuppliedtotheclassier.8.4.3SupportVectorMachineAnalysisNontransgenicNTvs.TransgenicTgGroupsError-ScoresOnly.AsshowninTable8.3,thesupportvectormachine-basedclas-siercorrectlyassigned88%Kappa=0.75ofcontrolnontransgenicandAlzheimer'stransgenicmicetotheirrespectivegroups,usingonlybehavioralerror-scoresfromalltasksoftheinterferenceparadigm.Boththesensitivityandspecicityoftheclassierwere88%.Hence,equalproportionsofeachgroupwerecorrectlyidentiedbytheclassier.Response-LatenciesOnly.Usingonlybehavioralresponse-latencydata,theclas-sierdemonstrated81%accuracyKappa=0.63,with75%sensitivityand88%specicity,asreportedinTable8.3.ErrorsandLatencies.Whenbotherror-scoresandresponse-latencieswerepro-videdtotheclassier,75%oftheanimalswerecorrectlyassignedtotheirrespectivegroupsKappa=0.50.Thesensitivitywas63%andthespecicitywas88%.Thisclassierexhibitedaperformancebiasfavoringnontransgenicmice,withgreaterlike-lihoodofmisclassifyingAlzheimer'stransgenics.Summary:Thesupportvectormachine-basedclassierssuccessfullydistinguishedbetweennontransgenicandAlzheimer'stransgenicmiceusingerror-scoredata.How-ever,theuseofresponse-latencydata,eitherexclusivelyorincombinationwitherror-scores,resultedinamarkeddeclineofbothaccuracyandspecicityforthetransgenicanimals.229

PAGE 240

NontransgenicNTvs.Nontransgenic-TreatmentNT+GMCSFGroupsError-ScoresOnly.Thesupportvectormachine-basedclassiercorrectlyidenti-ed63%ofindividualsKappa=0.25,with63%sensitivityand63%specicity,asshowninTable8.3.Response-LatenciesOnly.Theaccuracyoftheclassierincreasedto69%Kappa=0.38whenonlyresponse-latencydatafromtheinterferenceparadigmwereusedtodistinguishbetweenGMCSF-treatedanduntreatednontransgenicmice.Theob-servedsensitivitywas63%andthespecicitywas75%.Three-fourthsoftheun-treatednontransgenicswerecorrectlyidentied.ErrorsandLatencies.Trainingwithbotherror-scoresandresponse-latenciesresultedinaclassierwhichexhibitedonly56%accuracyKappa=0.13,asreportedinTable8.3.Thisclassiershowed63%sensitivityand50%specicity,erroneouslyidentifyinghalfoftheuntreatedanimalsashavingreceivedGMCSFtreatment.Summary:Overall,behavioralresponse-latencymeasuresprovidedthebestcri-teriafordistinguishingbetweenGMCSF-treatedanduntreatednontransgenicmiceusingsupportvectormachine-basedclassiers,althoughthediscriminationprovidedwasmodestatbest.Classierperformancedegradationoccurred,however,whenerror-scoreswereusedexclusivelyorincombinationwithresponse-latencies.TransgenicTgvs.Transgenic-TreatmentTg+GMCSFGroupsError-ScoresOnly.AsreportedinTable8.3,supportvectormachine-basedclas-sierstrainedexclusivelywithbehavioralerror-scoredatashowed75%accuracyKappa=0.50,with100%sensitivityand50%specicitynonsignicant.Hence,allGMCSF-treatedtransgenicswerecorrectlyidentied,whileone-halfofuntreatedanimalswereerroneouslyrecognizedashavingreceivedGMCSFtreatment.230

PAGE 241

Response-LatenciesOnly.Classierstrainedonlywithbehavioralresponse-latencydataexhibited75%overallaccuracyKappa=0.50,with75%sensitivityandspeci-city.Hence,individualsfrombothgroupswereequallylikelytobemisclassied.ErrorsandLatencies.Whenbotherrorsandlatencieswereusedtotraintheclassier,theaccuracyincreasedto82%Kappa=0.63,asindicatedinTable8.3,with100%sensitivityand63%specicity.Similartotheresultsobtainedfromusingerror-scoresexclusively,allGMCSF-treatedAlzheimer'stransgenicmicewerecorrectlyidentied.Summary:Supportvectormachine-basedclassierstrainedwitheithererror-scoreorresponse-latencydatawerecomparable,withrespecttooveralldiscrim-inabilitybetweenGMCSF-treatedanduntreatedtransgenicmice,thecombinationofbothbehavioralmeasuressignicantlyincreasedoverallperformancewhilepreserv-ingtheobservedsensitivityoftheError-ScoresOnlyclassiertotreatmenteectintransgenicanimals.AllFourGroupsError-ScoresOnly.Supportvectormachine-basedclassiersfailedtodistinguishamongallfourgroupsofmiceusingonlyerror-scoredatafromtheinterferenceparadigm.Indeed,onlyone-halfoftheuntreatedAlzheimer'stransgenicanimals,andnoneoftheuntreatednontransgenics,werecorrectlyidentied.Response-LatenciesOnly.Whenresponse-latencydatawereusedtotraintheclassier,theoverallaccuracywasapoor,albeitsignicant,28%Kappa=0.04,asshowninTable8.3.However,noneoftheGMCSF-treatedAlzheimer'stransgenicmiceoruntreatednontransgenicswerecorrectlyidentied.Only63%oftheuntreatedtransgenicanimalswerecorrectlyclassied.ErrorsandLatencies.Theclassierdidnotsuccessfullydistinguishamongthefourgroupsofanimalsusingbotherrorandlatencymeasuresfromtheinterfer-231

PAGE 242

enceparadigm.SimilartotheResponse-LatenciesOnlyresult,neithertheGMCSF-treatedtransgenicsnortheuntreatednontransgenicmicewererecognized,andonly63%ofuntreatedtransgenicswereidentiedcorrectly.Inaddition,fewerthanhalfoftheGMCSF-treatednontransgenicswerecorrectlyassignedtotheirtreatmentgroup.Summary:Overall,supportvectormachine-basedclassiersweregenerallyunsuc-cessfulindistinguishingamongthefourgroupsofmice.Signicantdiscriminabilitywasobservedwhenbehavioralresponse-latencydatawereusedexclusivelyaspredic-torvariables,buttheresultingclassiershowedexclusivebiasfavoringtheuntreatedAlzheimer'stransgenicmice,andwasotherwiseineective.8.5DiscussionStandardstatisticalanalysesusingANOVAofnal-blockbehavioralmeasuresindicatesignicantgroupwisedierencesinboththepost-testRadialArmwatermazeworkingmemorytrialT5error-score,aswellasinthreeofthefourtasksoftheinterferenceparadigmasshowninFigures8.1and8.2.ThesedierenceswereobservedbetweenTganimalsandtheotherthreegroups.Additionally,nontrans-genicmicebothNTandNT+GMCSFandGMCSF-treatedAlzheimer'strans-genicmicedidnotdiersignicantlyincognitiveperformanceintheRAWMtrialT5measure.Theseresultsareconsistentwithpriorndingse.g.,Leightyetal.,2004;Arendashetal,2006showingsignicantp<.05workingmemoryimpair-mentRAWMT4andT5measuresinAlzheimer'stransgenic-controlTgmice,relativetonontransgenic-controlsNT.Signicantdierencesinnal-blockper-formanceinboththethree-trialrecallanddelayed-recalltasksoftheinterferenceparadigmsuggestcognitiveimpairmentinAlzheimer'stransgenic-controlmice,rel-ativetonontransgenic-controls.Moreover,GMCSF-treatedAlzheimer'stransgenicmiceperformedsignicantlybetterinthedelayed-recalltaskduringthenalblock232

PAGE 243

oftheinterferenceparadigm,relativetotransgenic-controlanimals,indicativeofatherapeuticeectofGMCSFtreatment.Bycontrast,nosignicantdierenceswerefoundbetweenGMCSF-treatedanduntreatednontransgenicmiceinanyofthecomponentmeasuresoftheinterferenceparadigm,suggestingthatGMCSFmaynotprovidebenetsinindividualsnotpredisposedforAlzheimer-likepathology.Finally,althoughonlyasinglenon-interferenceparadigmmeasurewascomparedamongthegroups{nal-blockRAWMtrialT5error-scores{thecomponenttasksoftheinter-ferenceparadigmprovidesuperiorgroupwisediscriminabilityusingstandardstatis-ticalmethodsofanalysis.Extensivesignicantpairwisecorrelationwasobservedamongbotherror-scoreandresponse-latencymeasuresofallfourbehavioraltasks.Correspondingerrorandlatencymeasuresforeachtaskwerealsostronglycorrelated,underscoringthecomparabilityofthesebehavioralindicesforevaluatingcognitiveperformance.Bothbehavioralmeasuresoftheproactiveinterferencetaskweresignicantlycorrelatedwiththecorrespondingmeasureerrororlatencyofeachoftheothertasks,exceptforthree-trialrecall.ThesecorrelationresultsweresimilartothoseobtainedintheGRK5-relatedinvestigationrefertoChapter7,withtheexceptionoftheProactiveInterferencetaskcorrelationswithothertasks.Exploratoryfactoranalysisprovidedevidenceforsegregationbylearningcondi-tionmeasuresrelatedtoPoolAvs.measuresassociatedwithPoolBinthepre-cedingTheInterferenceTask:ANovelAssessmentParadigm"study,althoughsegregation-by-conditionwasnotobservedinthisstudy.Indeed,learningassociatedwithbothPoolAandPoolBwasreectedinthesecondfactor.Notsurprisingly,inlightoftheextensivepairwiseintercorrelationsbetween/amongmeasuresoftheThree-TrialRecallandDelayed-Recalltasks,thesetwocomponentsoftheinterfer-enceparadigmtogethercomprisedtheprimaryfactorTable8.2.Consequently,the233

PAGE 244

primaryfactormaybeinterpretedasrepresentingperformanceintheinitiallearn-ingtaskThree-TrialRecallandfollowingasubstantialdelayDelayed-Recall,thusreectingsuccessfulconsolidationofearlylearningPoolAintheinterfer-enceparadigmreminiscentoftheprimacyeectobservedinhumanserialrecall.Similarly,signicantintercorrelationbetween/amongmeasuresoftheProactiveIn-terferenceandRetroactiveInterferencetasksisreectedinthecoexistenceofthesetwotasksinFactorII,incontrasttothefactorstructureoftheinterferenceparadigmrecoveredintheprecedingstudywhereinonlyProactiveInterferencemeasurescom-prisedthesecondfactor.Thesecondfactormayberelatedtocontext-switchingorthecapacitytoadapttonewlearningconditionsPoolAvs.PoolB.Itisim-portanttorecognize,however,thatbothcorrelationandfactorstructuresmaybedierentifeitherrst-blockoroveralldatawereusedfortheseanalysesinsteadofnal-blockmeasures.Thismayoccur,inpart,asaconsequenceofthenonlin-eartime-dependenceoflearningandmemorye.g.,Gallisteletal.,2004,wherebydiscrepanciesingroupwiseanalyticresultsmayarisefrombehavioralsamplingatdierentstagesoflearning/memoryacquisition.Moreover,distinctiveoften,sub-tlestatisticalfeaturesofthesubjectpoole.g.,signicantlyunequalgroupsizes,dierentialtreatmenteect,discrepanciesingroupvariances,etc.mayintroducebiasorotherundesirableside-eectsintheanalysis.Thediscriminantanalysisresultsweregenerallyconsistentwithpriorstudies,whereinstepwise-forwardanalysesoftenreturnsuperiordiscriminabilityrelativetothestandarddirect-entrycompleteapproache.g.,ArendashandKing,2002;Leightyetal.,2004.Discriminantanalysis-basedclassierssuccessfullydistinguishedbothtransgenicityTgvs.NTandGMCSF-treatmenteectinAlzheimer'stransgenicmiceTgvs.Tg+GMCSF,andshowedmoderatediscriminability3%,albeitsuperiortoallotherclassiers,inthefour-groupcomparison.Indeed,despiteover-234

PAGE 245

allpoorperformance69%accuracy,thevariance-optimizingstepwise-forwardap-proachsuccessfullydistinguishedbetweentheNTandNT+GMCSFgroups,althoughtheconsensusamongclassierswasthatthesetwogroupsdonotdierwithrespecttocognitivemeasuresintheinterferenceparadigm.Thedecisiontree-basedclassierssuccessfullydistinguishedbetweennontrans-genicandAlzheimer'stransgenicmiceatalevelcomparabletodirect-entrydiscrim-inantanalysis,usinginterferencetaskerror-scoreswithorwithoutresponse-latencydata.Inaddition,decisiontreesdistinguishedbetweenGMCSF-treatedandun-treatedAlzheimer'stransgenicmiceatalevelcomparabletodirect-entrydiscrimi-nantanalysis.Incontrasttodiscriminantanalysis,theinclusionofresponse-latenciestotheerrordatadidnotimprovetheoverallaccuracyofthedecisiontree.Overalldiscriminabilityamongthefourgroupsofanimalsdemonstratedbydecisiontrees,usinginterferencetaskerror-scoresalone,wasbetweenthatofthetwodiscriminantanalysis-basedtechniques.NeuralnetworkswereparticularlyeectivefordistinguishingbetweenGMCSF-treatedanduntreatedAlzheimer'stransgenicmiceusingerror-scoresalone.However,neuralnetworksperformedcomparablytostepwise-forwarddiscriminantanalysiswhenusinglatencydatafordetectingthemaineectofAlzheimer'stransgenicityand,indeed,somewhatbetterthandecisiontreesfordistinguishingbetweenNTandNT+GMCSFindividuals.Supportvectormachinesperformedcomparablytodirect-entrydiscriminantanal-ysisfordistinguishingtransgenicityusingeithererror-scoreorresponse-latency,butnotboth,behavioralmeasuresfromtheinterferenceparadigm.Whencomparingbe-tweentheTgandTg+GMCSFgroups,however,thesupportvectormachine-basedclassierwaspoorerthandiscriminantanalysisandmarkedlyinferiortoneuralnet-workimplementations.Moreover,supportvectormachinesshowedthepoorestover-235

PAGE 246

alldiscriminabilityamongthefourgroups,achievinganacceptableperformancelevel5%orbetteronlyforresponse-latencydata.Inadditiontocomparisonsbyclassier-type,todeterminetherelativestrengthsandweaknessesofeachmethodology,itisinformativetocompareperformancebysubjectgroup,examiningbothpairwisecontrastsandthecompositefour-groupre-sults.Thisstudyexaminedtwomaineects,Alzheimer'stransgenicityandGMCSFadministration,inaclassic2x2design.Inaddition,twocognitivemetricserror-scoreandresponse-latencyweremeasuredforeachbehavioraltaskcomponentoftheinterferenceparadigm.ThemaineectofAlzheimer'stransgenicitywasbestde-tectedbystepwise-forwarddiscriminantanalysis,resultingin94%overallaccuracy,althoughtheotherclassiersalsoperformedremarkablywell.Additionally,classi-cationusingerror-scoresalonewasgenerallysuperiortoeortsincludingresponse-latencieswhichprovidednosignicantadvantage.Asdiscussedpreviously,theAPPgenotypeproducesadistinctbehavioralphenotype,withrespecttoprogressivecognitiveimpairment,andrepresentsausefulmodelforinvestigatingAlzheimer-likeneuropathologythroughitsovertbehavioralmanifestations.Indeed,standardsta-tisticalanalysesANOVAofbothnal-blockRAWMtrialT5error-scoresandtwotasksoftheinterferenceparadigmThree-TrialRecallandDelayed-Recall,bother-rorsandlatenciesclearlydierentiatedbetweenTgmiceandtheothergroups.TheconsensusamongclassiersisthattheGMCSF-treatmenteectinnontransgenicanimalsisminimal,i.e.,thesemicedonotbenetsignicantlyfromreceivingGM-CSFtreatment.Indeed,onlymodestdiscriminabilitywasconsistentlyobservedinbothneuralnetworksandsupportvectormachinesforthesetwogroups.ANOVA-basedanalysesofthenal-blockRAWMtrialT5failedtodistinguishbetweentheNTandNT+GMCSFgroups,andmodestbutnonsignicantdierenceswereob-servedinbotherror-scoresandresponse-latenciesoftheProactiveInterferenceand236

PAGE 247

RetroactiveInterferencetasksoftheinterferenceparadigm.TheGMCSF-treatmenteectinAlzheimer'stransgenicmice,bycontrast,issignicant.Indeed,allclassi-ersdistinguishedthistreatmenteectusingerror-scoresand/orresponse-latenciesobtainedintheinterferenceparadigm.Infact,onlydecisiontreeswereunabletodistinguishbetweenthegroupsusingonlylatencies.Standardstatisticalanalysisofnal-blockRAWMtrialT5showedasignicantimpairmentintheTggroup,relativetoTg+GMCSF.Similarly,ANOVA-basedanalysesofbothThree-TrialRecallandDelayed-RecallmeasureserrorsandlatenciesunderscorethecognitivedecitsofAlzheimer'stransgenic-controlmicerelativetotransgenicsreceivingGMCSFtreat-ment.Moreover,theTgandTg+GMCSFgroupswereoftenmoreeasilydistinguish-ablethanweretheTgandNTgroups.Error-scoresweregenerallymoreeectiveforachievingcompositefour-groupdiscriminability,althoughthestepwise-forwarddiscriminantanalysisprovidedthebestoverallperformanceusingacombinationoferrorsandlatencies%.Severalimportantconclusionsandcautionsmaybedrawnfromtheresults.First,theAlzheimer'stransgenicgenotypeAPPreliablyproducesadistinctivecognitive-behavioralphenotypeagainstwhichdiversegeneticmanipulationsandtherapeuticinterventionsmaybeexplored,analyzed,andinterpretedinthecontextofhumanAlzheimer'sdiseasediagnosis,treatment,andmanagement.Indeed,thecognitiverepertoireofthisbehavioralphenotypeissucientlyrich,withrespecttodiscernibleandquantiabledomainse.g.,workingmemory,thatcomprehensive,multitask-multimetricassessmentreturnsanextraordinarilydetailedportraitofabilities,aspreviouslyreportede.g.,Leightyetal.,2004.Second,thedierentialsensitivityofclassierstocognitive-behavioralfeaturessuggestsnotonlythesalienceofparticularattributesindividualmeasures,suchasthree-trialrecallerror-score,withrespecttogroupwisediscriminability,but,moreover,thelikelihoodthatnonlinearinteractions237

PAGE 248

between/amongattributescontributetotheobserveddiscriminability.Analysisofvariance,forexample,althoughwidelyacceptedinneurobehavioralresearch,displayssignicantlimitationswithrespecttogroupwisediscriminabilitye.g.,Figure8.1,dis-cussedearlierinpartbecauseofitsdependenceuponspecic,andoftenunrealistic,measurementconditions.Indeed,manystatisticalmethodologiesaredependentupon{andoptimizedfor{linearmodelsofpredictorvariablese.g.,ANOVA,multipleregression,lineardiscriminantanalysis.However,nonlinearinteractionsarequitecommoninnaturalphysiologicalsystemse.g.,Glass,2001;Westburyetal.,2003,andrequirespecialmathematicaltreatmentforproperanalysis.Asdiscussedearlier,neuralnetworksareparticularlysuitableforapplicationsinwhichtherelationshipwithinandbetweensetsofvariablesmeasuresislikelytobenonlinearor,indeed,unknown.Fourth,thisstudyfurthersupportsadoptionofthemouse-basedinterfer-enceparadigmasastandardcomponentofneurobehavioralassessmentforresearchinAlzheimer'stransgenicmice.Thecomponenttasksoftheparadigmwereshowntocontributeuniqueinformationreectingcognitivestatusinindividuals,aswellasdiscriminativecriteriafordistinguishingamonggroupsbytransgenicand/ortreat-menteect.Finally,althoughthisstudydidnotaddressthetherapeuticwindowe.g.,rst-blockvs.nal-blockvs.overallecacy,durationofresponsepermanentvs.transient,dose-dependence,orage-dependenceofGMCSFecacy,inlightoftheobservedtherapeuticeectofGMCSFtreatmentinAlzheimer'stransgenicani-malsand,possibly,toalesserextentinnontransgenics,additionalinvestigationofGMCSF'spotentialroleinADtherapyisindicated.238

PAGE 249

CHAPTER9CONCLUSIONSAlzheimer'sdiseaseisamultifactorialdisorder,exhibitingaremarkablycomplexportraitofneuropathologicalphenomenaandcognitive/behavioralmanifestations.Contemporaryresearchmethodologiesmustnecessarilyaddressthesefeaturesinor-dertodevelopandrenevalidexperimentalmodelse.g.,transgenicanimals,com-prehensiveneurobehavioralassessmentparadigms,andrigorousdataanalysisproto-cols.Indeed,thesophisticationofmodeling,evaluation,andanalysiscloselyparallelsadvancesinbothcomputational-andbio-technology.Consequently,today'smedicalscientistsandclinicianspossessapowerfularrayoftoolsforinvestigatingbothpatho-logicalfrommolecular-genetictoorgan-systemscaleandbehavioralsensorimotor,cognitive,dailylivingaspectsofAlzheimer'sdisease.TheresearchpresentedinthisdissertationaddressedneurobehavioralassessmentandanalysisinbothAlzheimer'stransgenicmiceandhumanADpatients.Theutilityofcomprehensivesensorimotorandcognitivebehavioralevaluationforbothtrans-genicandtreatmenteectswasdemonstrated,underscoringtheimportanceofmulti-pletasksand/ormeasuresforidentifyingandquantifyinggroup-dierencesinanimalmodelsofhumanAD.Dataminingtechniquesdecisiontrees,neuralnetworks,sup-portvectormachines,underrepresentedinbehavioralresearch,werecomparedwithconventionalstatisticalanalysescorrelation,discriminantanalysis,factoranalysistoexplorethediagnosticpotentialofdataminingmethodologiesinbothhuman-andmouse-basedstudies.Asemanticinterferencelearningtaskoriginallydeveloped239

PAGE 250

forhumanADassessmentwasadaptedformice,byreplacingverbal-semanticele-mentswithvisual-spatialcomponents.Theresultingparallelevaluationparadigmeectivelydistinguishedbothtransgenicandtreatmenteectsinanimalmodels.Comprehensiveassessment,usingmultipletasksandmeasuresofsensorimotorandcognitiveability,providesthebestopportunitytodetectanomalousbehav-iorssuggestiveofunderlyingneuropathology.Bysamplingacrossdiversebehavioraldomainse.g.,balance,agility,spatiallearning,referencememoryandrecordingmultipleintrataskresponsemetricse.g.,acquisitionandretentionmeasuresintheMorriswatermazetask,abroadercross-sectionofthebehavioralrepertoireisre-vealed,thusprovidingtheinvestigatorwithamorecompleteportraitofthesubject'soverallneurobehavioralstatus.Subsequentanalysisofbehavioralmeasuresrevealsevidenceofthecomplexinterplayamong,aswellaswithin,sensorimotorandcog-nitivedomainswhich,collectively,areexpressedasadistinctsyndrome.Thesig-nicantcorrelationsbetween/amongbehavioralmeasuresobservedwithinindividualcognitive-loadedtasks,suchastheradialarmwatermaze,reectsboththeextentofintegrationamongdiversecognitivefunctionsnecessaryforsuccessfuloverallper-formanceinthesetasks,aswellasthesweepingimpactofincipientneuropathologyacrossmultiplecognitivedomains.Strongconsensusamongcognitivemeasuresob-tainedfromthewatermaze-typetasksRAWM,Morriswatermaze,platformrecog-nitiontaskunderscorestheabilityoftheseprotocolstoevaluatesimilarfeaturesoflearningandmemoryprocesses.Indeed,factoranalysesofthecomprehensivetaskbatteryconsistentlyreturnedaprimaryfactorcomprisedlargelyofmeasuresfromtheRAWMandplatformrecognitiontasks,representingoverallcognitivefunction-ing,whichmaybeanalogoustotheputativegeneralmentalabilityg"constructinhumans.240

PAGE 251

Statisticalapproachesforidentifyingandquantifyinggroupdierencesformtheanalyticfoundationofbothexperimentalandclinicalneurobehavioralresearch.Dataminingtechniques,bycontrast,havebeenconnedlargelytoindustrialandcommer-cialapplications,andareonlyrecentlygainingrecognitioninbiomedicalresearch.Classiermodelsproducedbystepwise-forwarddiscriminantanalysisweregenerallysuperiortothosereturnedbydirect-entrycompletediscriminantanalysisfordistin-guishingbetween/amonggroupsbasedontreatmentand/ortransgeniceects,usingbehavioralmeasures.Thevariance-optimizationcapacityofthestepwise-forwardap-proach,inconjunctionwiththeavailabilityofarelativelylargesamplingofthemousebehavioralrepertoirei.e.,manymeasuresfromwhichtoselectpredictorvariables,increasedthelikelihoodforsuccessfuldiscriminability.Theresultingclassiersoftenselectedacollectionofsensorimotor,anxiety,andcognitivemeasures,stressingtheimportanceofmultimetricassessmentinventories.Moreover,thesendingsunder-scoretheutilityoftheassessmentbatteryformousebehavioralphenotypingchar-acterizinggroupsofanimalsbytreatmentand/ortransgeniceectsusingdiscernablebehavioralfeatures.Decisiontrees,whichselectpredictorvariablesbasedoninformation-biasca-pacity,didnotexhibitconsistentperformancebut,rather,idiosyncraticsubjecttodierentialperformancewithdierentdatasetsbehaviorintherelatively-smallerdatasetsofmouse-basedstudies.Forthisreason,unlesstheinvestigatorhassubstan-tialpriorexperiencewiththeexperimentaltreatmentconditions,tomakeinformeddecisionsconcerningpredictorvariableselection,relianceondecisiontree-basedclas-siersisnotrecommended,particularlywhensmallsamplesizesarepresent.Neu-ralnetworksandsupportvectormachinesalsoexhibitedmodestlyidiosyncratic,albeitmoreconsistent,behaviorforanalyzinggroupwisebehavioralmeasures.In-deed,either/bothofthesetwomethodswerefoundtobesuperiortodiscriminant241

PAGE 252

analyses,withrespecttogroupwisediscriminability.Thesendingssuggestthatmemory/learningprocesses,mirroringtheunderlyingphysiologicalsubstrates,mayexhibitnonlinearbehavioralmanifestationstowhichneuralnetworkandsupportvectormachinearchitecturesmaybemoresensitivecomparedtomultivariatesta-tisticalanalyses.Moreover,theseresultsindicatethatstatistical-anddatamining-basedclassiersmaybecomplementarytechniques,ratherthansubstitutesand,inaddition,thatsamplesizemaybeanimportantconsiderationwhenselectingamongalternativeanalyticmethodologies.Twostudiesexaminingthecognitiveeectsofcaeineadmininstrationinmicewereusedtodevelopandreneadvancedstatisticalanddatamininganalytictech-niquesforneurobehavioralresearchinbothnontransgenicandAlzheimer'strans-genicanimals.TherststudyChapter4,involvingagednontransgenicmicewhichreceivedoralcaeine.5mg/dayforapproximatelytenmonthssuggeststhat,al-thoughcaeinemayexertsubtleinuenceonbothsensorimotorandcognitivebe-havior,itisunlikelythatchroniccaeineintakeprovidescognitivebenetsinnormalnontransgenicanimals,asreectedinpoorgroupwisediscriminabilitybybothad-vancedstatisticalanddatamining-basedclassiers.Theeectsoflong-termcaeineadministrationinAlzheimer'stransgenicmicewereinvestigatedinthesecondstudyChapter5.Afterconsumingoralcaeine.5mg/dayforfourtoveweeks,caeine-treatedtransgenicanimalsperformedcomparablytoage-matchednontrans-genicmiceand,indeed,superiortountreatedtransgenicanimals,withrespecttocognitivemeasuresofworkingmemory,asindicatedbysuperiordiscriminabilitybe-tween/amongtreatmentgroupsbybothsetsofclassiers.ThesemanticinterferencetaskLoewensteinetal.,2004wasrecentlydevelopedasaclinicaldiagnostic/screeninginstrumentforprobableAlzheimer'sdisease.Theoriginalhumanversionconsistsofverbally-reportedimmediate-anddelayed-recall242

PAGE 253

memorycomponents,whichassessthesubject'sabilitytomaintaintwodistinctcol-lectionsoffamiliaritemsinmemory.WhenthedatasetfromtheLoewensteinetal.004studywasanalyzedusingadvancedstatisticalanddatamining-basedtech-niquesChapter6,thesemethodologieswerefoundtobecomparablewithrespecttooverallaccuracy,sensitivity,andspecicityfordistinguishingbetween/amongprob-ableAlzheimer'sdisease,mildlycognitively-impaired,andnormalagedindividualshumans.Collectively,theseresultsunderscorethediagnosticimportanceofthesemanticinterferenceparadigmforAlzheimer'sscreening,aswellastheuseofdataminingtechniquesforclinicalapplicationstypicallyconnedtostatisticalmethods.Amouse-basedadaptationofthetaskwascreatedbysubstitutingspatiallearn-ingelementssimilarapparatustotheRAWMfortheverbalcomponentsoftheoriginaltask.Theresultingparadigmgeneratedperformancemeasuresbotherror-scoresandresponse-latencieswhichwereanalogoustothoseoftheoriginalhuman-baseddesign,reecting:generalrecallandlearningcapacity,proactiveinterference,retroactiveinterference,anddelayed-recallability.Themouse-basedinterferenceparadigmwascomparedtoconventionalneurobe-havioralassessmentprotocolse.g.,RAWM,comprehensivebehavioraltaskbatteryintwostudies,involving:thecognitiveeectsofanadditionalgeneticmanip-ulationeliminationofGRK5expressionsuperimposedontheAlzheimer'strans-genicgenotype,aswellasnontransgenics,and,2therapeuticecacyofGM-CSFadministrationinAlzheimer'stransgenicmiceandage-matchednontransgenican-imals.TheGRK5-knockoutgenotypeisbeinginvestigatede.g.,Suoetal.,2007asamodelforAlzheimer's-likeneuropathology.InChapter7,one-year-oldmicebothnontransgenicandAlzheimer'stransgenicwithanadditionalgeneticmodi-cationwith/withoutGRK5expressioncompletedboththecomprehensivebehav-ioraltaskbatteryandthemouse-basedinterferenceparadigm,andtheresultswere243

PAGE 254

analyzedusingstandardstatisticsANOVA,advancedstatisticse.g.,discriminantanalysis,anddataminingmethodsdecisiontrees,neuralnetworks,supportvectormachines.NontransgenicandAlzheimer'stransgenicmice,bothexpressingGRK5,werefoundtodierinbothsensorimotorandcogitivemeasuresofthecomprehen-sivebattery,usingbothstandardandadvancedstatistics,aswellasdataminingmethods.Thesetwogroupsalsoexhibitsignicantdierencesincomponentmea-suresoftheinterferenceparadigm,basedonstandardandadvancedstatistics,aswellasdatamininganalyses.Bycontrast,onlymodestdiscriminabilitywithrespecttoGRK5expressionwasobservedinbothnontransgenicandAlzheimer'stransgenicanimals.However,comparisonsamongnontransgenicmiceexpressingGRK5andbothAlzheimer'stransgenicgroupsacrossmeasuresintheinterferenceparadigmsuggestthateliminatingexpressionofGRK5mayamelioratecognitiveimpairmentassociatedwiththeAlzheimer'stransgene,bringingtheTg-KOanimalsclosertothebaselinecognitiveperformanceofGRK5-expressingnontransgenics.Indeed,GRK5-expressingnontransgenicsonlydiersignicantlyfromGRK5-knockoutAlzheimer'stransgenicswithrespecttoasinglebehavioralmeasureoftheinterferenceparadigm,retroactiveinterferenceresponse-latency.Finally,thetherapeuticecacyofGM-CSFadmininstration,withrespecttocognitivefunction,wasexaminedinbothnon-transgenicandAlzheimer'stransgenicanimalsChapter8.TheinammatogeniccytokineGM-CSFhasbeenproposedasatreatmentforAlzheimer'sdisease,toatten-uatebeta-amyloidplaque-associatedstructuralandfunctionalimpairment,therebyimprovingorrestoringcognitivefunction.One-year-oldAlzheimer'stransgenicmicereceiveddailyinjectionsofGM-CSFmcg/dayfortwoweeks,thencompletedbothRAWMpost-testingandtheinterferenceparadigm.GMCSF-treatedtransgenican-imalsperformedsigncantlysuperiortountreatedtransgenicsandcomparablytoage-matchednontransgenicmicereceivingeitherGMCSForplainvehicleinthe244

PAGE 255

RAWMtrialT5workingmemoryerror-scoremeasure,aswellasinthethree-trialrecallanddelayed-recallmeasuresoftheinterferenceparadigm.GMCSF,therefore,mayameliorateAlzheimer'stransgene-associatedcognitiveimpairmentand,thus,representacognitive-protectivetreatmentforindividualspredisposedforAlzheimer'sdisease.Intheory,earlydetectionandtherapeuticinterventionrepresentsthebeststrat-egyfortreatingprogressivedisorders,suchasAlzheimer'sdisease.Althoughspe-cicgeneticmarkershavebeenidentiedasriskfactorsforearly-onset,familialAlzheimer'sdisease,theetiologyofmostADcasesoverninetypercentremainsunclear.Manysituationale.g.,lifestyle,dietanddispositionale.g.,otherge-neticfactors,epigeneticscandidateriskfactorsarecurrentlyunderinvestigation.Meanwhile,diagnostic,therapeutic,andmanagementprotocolscurrentlyexistforpatients.Subtlemanifestationsofnascentcognitiveimpairment,forexample,mayberevealedthroughearlypsychometricindicese.g.,academicormentalabilitytestsubscores.Certainindividualdierencesinlearningpatternse.g.,memory-tasklearningcurvewhicharecommonlydismissedaspsychologicalvariabilitymay,infact,representveryearlyindicationsofcognitivedysfunctionor,possibly,pre-dispositiontodementia-likesyndromes.Advancedanalyticmethodologies,aswell,mayleadtonewinsightsthroughtheintegrationofneurobehavioralandneurologicaldatabases.Customizedhybridandensembleclassiersystems,forinstance,couldbeusedtodetectspecicdistinguishingfeatures,patterns,ortrendswhichreliablyiden-tifypathologicstates,therapeuticecacy,ordierentialtreatmenteectsinbothexperimentale.g.,transgenicanimalsandclinicalsettings.Finally,inrecognitionofthespatialandtemporalaspectsofAlzheimer'sdisease,investigativeparadigmswhichexploitthespatiotemporaldynamicsofADshouldbepursued.Forexample,certaincomputationalarchitectures,suchasself-organizingmaps,areparticularly245

PAGE 256

suitableformultidimensionalmodelingofnonlineardynamicalprocesses,suchasregiospecicpatternsofbrainactivation.ContemporaryAlzheimer'sresearchhasbecomeamultidisciplinaryendeavor,bringingtogetherthediverseinsightsandtalentsofbothacademicandprofessionalspecialties.Alzheimer'sdiseaserepresentsasignicantpublichealthconcernwhichcallsforacommensurateshareofsocial,political,andscienticattention.Indeed,thereisgrowingpublicawarenessofAlzheimer'sdiseaseandrelatedresearchactivi-ties,asreportedbythenewsmediaandportrayedinpopularentertainment.Eachadvancerenewsourhopes,andeachsetbackstrengthensourresolve:Werecognizethisfamiliarebbandowasthenaturalpaceofprogress,andndsolaceinitscontinuity.246

PAGE 257

REFERENCESAbdi,H.94.Aneuralnetworkprimer.JournalofBiologicalSystems,2,247-283.Ackley,D.H.,Hinton,G.E.,andSejnowski,T.J.85.AlearningalgorithmforBoltzmannmachines.CognitiveScience,9,147-169.Adlard,P.A.,andCummings,B.J.04.Alzheimer'sdisease{Asumgreaterthanitsparts?NeurobiologyofAging,25,725-733.Adrian,E.D.,andYamagiwa,K.935.TheoriginoftheBergerrhythm.Brain,58,323-351.Aftanas,L.I.,andGolocheikine,S.A.02.Non-lineardynamiccomplexityofthehumanEEGduringmeditation.NeuroscienceLetters,330,143-146.Aggarwal,B.B.,andShishodia,S.04.SuppressionofthenuclearfactorBactivationpathwaybyspice-derivedphytochemicals:Reasoningforseasoning.AnnalsoftheNewYorkAcademyofScience,1030,434-441.Alkon,D.L.974.AssociativetrainingofHermissenda.JournalofGeneralPhysiology,64,70-84.Almeida,J.S.2002.Predictivenon-linearmodelingofcomplexdatabyarticialneuralnetworks.CurrentOpinioninBiotechnology,13,72-76.Alonso,A.delC.,Zaidi,T.,Novak,M.,Grundke-Iqbal,I.,andIqbal,K.001.Hyperphosphorylationinducesself-assemblyoftauintotanglesofpairedhelicallaments/straightlaments.ProceedingsoftheNationalAcademyofSciencesUSA,98,6923-6928.Altman,D.G.,andBland,J.M.94.Diagnostictests.1.Sensitivityandspecicity.BritishMedicalJournal,308,1552.Alzheimer,A.907.UbereineeigenartigeErkrankungderHirnrinde.AllgemeineZeitschriftfurPsychiatrieundpsychiatrisch-gerichtlicheMedizinBerlin,64,146-148.Alzheimer'sAssociation.08.2008Alzheimer'sDiseaseFactsandFigures.Alzheimer'sandDementia,42.247

PAGE 258

Alzheimer'sDiseaseCollaborativeGroup995.Thestructureofthepresenilin1S182geneandidenticationofsixnovelmutationsinearlyonsetADfamilies.NatureGenetics,11,219-222.Amari,S.,Beltrame,F.,Bjaalie,J.G.,Dalkara,T.,DeSchutter,E.,Egan,G.F.,Goddard,N.H.,Gonzalez,C.,Grillner,S.,Herz,A.,Homann,K.P.,Jaaske-lainen,I.,Koslow,S.H.,Lee,S.Y.,Matthiessen,L.,Miller,P.L.,DaSilva,F.M.,Novak,M.,Ravindranath,V.,Ritz,R.,Ruotsalainen,U.,Sebestra,V.,Subramaniam,S.,Tang,Y.,Toga,A.W.,Usui,S.,VanPelt,J.,Verschure,P.,Willshaw,D.,Wrobel,A.,andtheOECDNeuroinformaticsWorkingGroup.002.Neuroinformatics:Theintegrationofshareddatabasesandtoolsto-wardintegrativeneuroscience.JournalofIntegrativeNeuroscience,1,117-128.Ambree,O.,Leimer,U.,Herring,A.,Gortz,N.,Sachser,N.,Heneka,M.T.,Paulus,W.,andKeyvani,K.06.ReductionofamyloidangiopathyandAbetaplaqueburdenafterenrichedhousinginTgCRND8mice.AmericanJournalofPathology,169,544-552.Amieva,H.,Lafont,S.,Rouch-Leroyer,I.,Rainville,C.,Dartiques,J.F.,Orgogozo,J.M.,andFabriqoule,C.004.EvidencinginhibitorydecitsinAlzheimer'sdiseasethroughinterferenceeectsandshiftingdisabilitiesintheStrooptest.ArchivesofClinicalNeuropsychology,19,791-803.Amouyel,P.,Vidal,O.,Launay,J.M.,andLaplanche,J.L.994.Theapolipopro-teinEallelesasmajorsusceptibilityfactorsforCreutzfeldt-Jakobdisease.Lancet,344,1315-1318.Angelucci,M.E.,Vital,M.A.,Cesario,C.,Zadusky,C.R.,Rosalen,P.L.,andDuCunha,C.99.Theeectofcaeineinanimalmodelsoflearningandmemory.EuropeanJournalofPharmacology,373,135-140.Arendash,G.W.,Garcia,M.F.,Costa,D.A.,Cracchiolo,J.R.,Wefes,I.M.,andPotter,H.004a.EnvironmentalenrichmentimprovescognitioninagedAlzheimer'stransgenicmicedespitestable-amyloiddeposition.NeuroReport,15,1751-1754.Arendash,G.W.,Jensen,M.T.,Salem,N.,Jr.,Hussein,N.,Cracchiolo,J.,Dick-son,A.,Leighty,R.,andPotter,H.2007.Adiethighinomega-3fattyacidsdoesnotimproveorprotectcognitiveperformanceinAlzheimer'stransgenicmice.Neuroscience,149,286-302.Arendash,G.W.,andKing,D.L.02.Intra-andintertaskrelationshipsinabe-havioraltestbatterygiventoTg2576transgenicmiceandcontrols.PhysiologyandBehavior,75,643-652.248

PAGE 259

Arendash,G.,King,D.,Gordon,M.,Morgan,D.,Hatcher,J.,Hope,C.,andDiamond,D.001.Progressive,age-relatedbehavioralimpairmentsintrans-genicmicecarryingbothmutantamyloidprecursorproteinandpresenilin-1transgenes.BrainResearch,891,42-53.Arendash,G.W.,Lewis,J.,Leighty,R.E.,McGowan,E.,Cracchiolo,J.R.,Hut-ton,M.,andGarcia,M.F.04b.Multi-metricbehavioralcomparisonofAPPswandP301LmodelsforAlzheimer'sDisease:Linkageofpoorercogni-tiveperformancetotaupathologyinforebrain.BrainResearch,1012,29-41.Arendash,G.W.,Mori,T.,Cao,C.,Mamcarz,M.,Runfeldt,M.,Dickson,A.,Rezai-Zadeh,K.,Tan,J.,Citron,B.A.,Lin,X.,Echeverria,V.,andPotter,H.009.CaeinereversescognitiveimpairmentanddecreasesbrainAlevelsinagedAlzheimer'smice.JournalofAlzheimer'sDisease.Inpress.Arendash,G.W.,Schleif,W.,Rezai-Zadeh,K.,Jackson,E.K.,Zacharia,L.C.,Cracchiolo,J.R.,Shippy,D.,andTan,J.06.CaeineprotectsAlzheimer'smiceagainstcognitiveimpairmentandreducesbrain-amyloidproduction.Neuroscience,142,941-952.Astur,R.S.,Tropp,J.,Sava,S.,Constable,R.T.,andMarkus,E.J.04.SexdierencesandcorrelationsinavirtualMorriswatertask,avirtualradialarmmaze,andmentalrotation.BehavioralBrainResearch,151,103-115.Atkinson,G.F.66.Designsforsequencesoftreatmentswithcarryovereects.Biometrics,22,292-309.Backman,L.,Jones,S.,Berger,A.-K.,Laukka,E.J.,andSmall,B.J.005.Cog-nitiveimpairmentinpreclinicalAlzheimer'sdisease:Ameta-analysis.Neu-ropsychology,19,520-531.Backman,L.,Small,B.J.,andFratiglioni,L.01.StabilityofthepreclinicalepisodicmemorydecitinAlzheimer'sdisease.Brain,124,96-102.Baddeley,A.D.92.Workingmemory.Science,255,556-559.Baddeley,A.D.,andHitch,G.J.74.Workingmemory.InG.H.BowerEd.RecentAdvancesinLearningandMotivation,8,47-90.NewYork:AcademicPress.Bai,D.L.,Tang,X.C.,andHe,X.C.00.HuperzineA:ApotentialtherapeuticagentfortreatmentofAlzheimer'sdisease.CurrentMedicinalChemistry,7,355-374.249

PAGE 260

Bales,K.,Verina,T.,Cummins,D.,Du,Y.,Dodel,R.,Saura,J.,Fishman,C.,DeLong,C.,Piccardo,P.,Petegnief,V.,Ghetti,B.andPaul,S.1999.ApolipoproteinEisessentialforamyloiddepositionintheAPPV171Ftrans-genicmousemodelofAlzheimer'sdisease.ProceedingsoftheNationalAcademyofSciencesUSA,96,15233-15238.Balestreri,R.,Fontana,L.,andAstengo,F.1987.Adouble-blindplacebocon-trolledevaluationofthesafetyandecacyofvinpocetineinthetreatmentofpatientswithchronicvascularsenilecerebraldysfunction.JournaloftheAmericanGeriatricSociety,35,425-430.Ba~nos,J.H.,andFranklin,L.M.002.FactorstructureoftheMini-MentalStateExaminationinadultpsychiatricinpatients.PsychologicalAssessment,14,397-400.Barnes,C.A.79.Memorydecitsassociatedwithsenescence:Aneurophysi-ologicalandbehavioralstudyintherat.JournalofComparativeandPhysio-logicalPsychology,93,74-104.Barnes,J.,Scahill,R.I.,Boyes,R.G.,Frost,C.,Lewis,E.B.,Rossor,C.L.,Rossor,M.N.,andFox,N.C.004.DierentiatingAlzheimer'sdiseasefromagingusingsemiautomatedmeasurementofhippocampalatrophyrates.Neuroimage,23,574-581.Bartlett,M.S.37.Propertiesofsuciencyandstatisticaltests.ProceedingsoftheRoyalSociety,SeriesA,160,268-282.Bartlett,M.S.1947.Theuseoftransformations.Biometrics,3,39-52.Bassett,S.S.,Avramopoulos,D.,andFallin,M.D.002.Evidenceforparentoforigineectinlate-onsetAlzheimerdisease.AmericanJournalofMedicalGeneticsNeuropsychiatricGenetics,114,679-686.Bassett,S.S.,Avramopoulos,D.,Perry,R.T.,Wiener,H.,Watson,B.,Go,R.C.P.,andFallin,M.D.06.Furtherevidenceofamaternalparent-of-origineectonChromosome10inlate-onsetAlzheimer'sdisease.AmericanJournalofMedicalGeneticsPartBNeuropsychiatricGenetics,141B,537-540.Bauer,E.,andKohavi,R.1999.Anempiricalcomparisonofvotingclassicationalgorithms:Bagging,boosting,andvariants.MachineLearning,36,105-142.Baxt,W.G.995.Applicationofarticialneuralnetworkstoclinicalmedicine.Lancet,346,1135-1138.Benzing,W.,Wujek,J.,Ward,E.,Shaer,D.,Ashe,K.,Younkin,S.,andBrunden,K.999.Evidenceforglial-mediatedinammationinagedAPPswtransgenicmice.NeurobiologyofAging,20,581-589.250

PAGE 261

Berman,K.,andBrodaty,H.004.TocopherolvitaminEinAlzheimer'sdiseaseandotherneurodegenerativedisorders.CNSDrugs,18,807-825.Biegler,R.,McGregor,A.,Krebs,J.R.,andHealy,S.D.01.Alargerhip-pocampusisassociatedwithlonger-lastingspatialmemoryProceedingsoftheNationalAcademyofSciencesUSA,98,6941-6944.Billings,L.M.,Oddo,S.,Green,K.N.,McGaugh,J.L.,andLaFerla,F.M.005.IntraneuronalAbetacausestheonsetofearlyAlzheimer'sdisease-relatedcognitivedecitsintransgenicmice.Neuron,45,675-688.Bjornsson,H.T.,Sigurdsson,M.I.,Fallin,M.D.,Irizarry,R.A.,Aspelund,T.,Cui,H.,Yu,W.,Rongione,M.A.,Ekstrom,T.J.,Harris,T.B.,Launer,L.J.,Eiriksdottir,G.,Leppert,M.F.,Sapienza,C.,Gudnason,V.,andFeinberg,A.P.008.Intra-individualchangeovertimeinDNAmethylationwithfamilialclustering.JournaloftheAmericanMedicalAssociation,299,2877-2883.Blacker,D.,Albert,M.S.,Bassett,S.S.,Go,R.C.,Harrell,L.E.,andFolstein,M.F.94.ReliabilityandvalidityofNINCDS-ADRDAcriteriaforAlzheimer'sdisease:TheNationalInstituteofMentalHealthgeneticsinitiative.ArchivesofNeurology,51,1198-1204.Blacker,D.,Bertram,L.,Saunders,A.J.,Moscarillo,T.J.,Albert,M.S.,Wiener,H.,Perry,R.T.,Collins,J.S.,Harrell,L.E.,Go,R.C.P.,Mahoney,A.,Beaty,T.,Fallin,M.D.,Avramopoulos,D.,Chase,G.A.,Folstein,M.F.,McInnis,M.G.,Bassett,S.S.,Doheny,K.J.,Pugh,E.W.,andTanzi,R.E.2003.Resultsofahigh-resolutiongenomescanof437Alzheimer'sdiseasefamilies.HumanMolecularGenetics,12,23-32.Bland,J.M.,andAltman,D.G.96a.Statisticsnotes:Transformingdata.BritishMedicalJournal,312,770.Bland,J.M.,andAltman,D.G.996b.Statisticsnotes:Transformations,means,andcondenceintervals.BritishMedicalJournal,312,1079.Bliss,T.,andLomo,T.73.Long-lastingpotentiationofsynaptictransmis-sioninthedentateareaoftheanesthetizedrabbitfollowingstimulationoftheperforantpath.JournalofPhysiology,232,331-356.Bloom,F.E.,Reilly,J.F.,Redwine,J.M.,Wu,C.-C.,Young,W.G.,andMorrison,J.H.005.Mousemodelsofhumanneurodegenerativedisorders:Require-mentsformedicationdevelopment.ArchivesofNeurology,62,185-187.Bock,C.,andLengauer,T.008.Computationalepigenetics.Bioinformatics,24,1-10.251

PAGE 262

Bohbot,V.D.,Jech,R.,Ruzicka,E.,Nadel,L.,Kalina,M.,Stepankova,K.,andBures,J.002.Ratspatialmemorytasksadaptedforhumans:Characteriza-tioninsubjectswithintactbrainandsubjectswithselectivemedialtemporallobethermallesions.PhysiologicalResearch,51Suppl.1,S49-S65.Bondi,M.W.,Serody,A.B.,Chan,A.S.,Eberson-Shumate,S.C.,Delis,D.C.,Hansen,L.A.,andSalmon,D.P.002.CognitiveandneuropathologiccorrelatesofStroopColor-WordTestperformanceinAlzheimer'sdisease.Neu-ropsychology,16,335-343.Boneau,A.960.Theeectsofviolatingassumptionsunderlyingthettest.PsychologicalBulletin,57,49-64.Borchelt,D.,Ratovitski,T.,vanLare,J.,Lee,M.,Gonzales,V.,Jenkins,N.,Copeland,N.,Price,D.,andSisodia,S.97.Acceleratedamyloiddeposi-tioninthebrainsoftransgenicmicecoexpressingmutantpresenilin1amyloidprecursorproteins.Neuron,19,939-945.Borras,D.,Ferrer,I.,andPumarola,M.1999.Age-relatedchangesinthebrainofthedog.VeterinaryPathology,36,202-211.Boser,B.E.,Guyon,I.M.,andVapnik,V.N.992.Atrainingalgorithmforoptimalmarginclassiers.In:Haussler,D.Ed.Proceedingsofthe5thAnnualACMWorkshoponComputationalLearningTheory,pp.144-152.ACMPress.Bossy-Wetzel,E.,Schwarzenbacher,R.,andLipton,S.A.04.Molecularpath-waystoneurodegeneration.NatureMedicine,10,S2-S9.Bourin,M.,Chenu,F.,Ripoll,N.,andDavid,D.J.P.2005.Aproposalofdecisiontreetoscreenputativeantidepressantsusingforcedswimandtailsuspensiontests.BehavioralBrainResearch,164,266-269.Bourre,J.M.04.Rolesofunsaturatedfattyacidsespeciallyomega-3fattyacidsinthebrainatvariousagesandduringaging.JournalofNutrition,HealthandAging,8,163-174.Box,G.E.P.,andCox,D.R.964.Ananalysisoftransformations.JournaloftheRoyalStatisticalSociety,SeriesB,26,211-252.Bracco,L.,Gallato,R.,Grigoletto,F.,Lippi,A.,Lepore,V.,Bino,G.,Lazzaro,M.P.,Carella,F.,Piccolo,T.,andPozzilli,C.994.FactorsaectingcourseandsurvivalinAlzheimer'sdisease:A9-yearlongitudinalstudy.ArchivesofNeurology,51,1213-1219.Brasnjevic,I.,Steinbusch,H.W.M.,andSchmitz,C.006.Alteredgeneexpres-sionandneuropathologyinAlzheimer'sdisease.NeurobiologyofAging,27,1081-1083.252

PAGE 263

Brenner,R.P.,Ulrich,R.F.,Spiker,D.G.,Sclabassi,R.J.,Reynolds,C.F.,Marin,R.S.,andBoller,F.86.ComputerizedEEGspectralanalysisinelderlynormal,demented,anddepressedsubjects.ElectroencephalographyandClinicalNeurophysiology,64,483-492.Brenner,R.P.,Reynolds,C.F.,andUlrich,R.F.988.DiagnosticecacyofcomputerizedspectralversusvisualEEGanalysisinelderlynormal,demented,anddepressedsubjects.ElectroencephalographyandClinicalNeurophysiology,69,110-117.Breslow,N.970.AgeneralizedKruskal-WallistestforcomparingKsamplessubjecttounequalpatternsofcensorship.Biometrika,57,579-594.Briley,M.,Chopin,P.,andVeigner,M.986.Theplusmazetestofanxiety,"validationindierentratstrainsandeectofawidevarietyofantidepressants.BritishJournalofPharmacology,87,217P.Britschgi,M.,andWyss-Coray,T.2007.SystemicandacquiredimmuneresponsesinAlzheimer'sdisease.InternationalReviewofNeurobiology,82,205-233.Broadhurst,P.L.1961.Analysisofmaternaleectsintheinheritanceofbehavior.AnimalBehavior,9,129-141.Brookmeyer,R.,Gray,S.,andKawas,C.1998.ProjectionsofAlzheimer'sdiseaseintheUnitedStatesandthepublichealthimpactofdelayingdiseaseonset.AmericanJournalofPublicHealth,88,1337-1342.Brouwers,N.,Sleegers,K.,Engelborghs,S.,Bogaerts,V.,Serneels,S.,Kamali,K.,Corsmit,E.,DeLeenheir,E.,Martin,J.-J.,DeDeyn,P.P.,VanBroeck-hoven,C.,andTheuns,J.06.GeneticriskandtranscriptionalvariabilityofamyloidprecursorproteininAlzheimer'sdisease.Brain,129,2984-2991.Brown,J.A.58.Sometestsofthedecaytheoryofimmediatememory.Quar-terlyJournalofExperimentalPsychology,10,12-21.Buchman,A.S.,Schneider,J.A.,Wilson,R.S.,Bienias,J.L.,andBennett,D.A.006.BodymassindexinolderpersonsisassociatedwithAlzheimerdiseasepathology.Neurology,67,1949-1954.Bures,J.,andBuresova,D.963.Corticalspreadingdepressionasamemorydisturbingfactor.JournalofComparativePhysiologyandPsychology,56,268-272.Burges,C.J.C.998.Atutorialonsupportvectormachinesforpatternrecog-nition.DataMiningandKnowledgeDiscovery,2,121-167.253

PAGE 264

Burn,D.,Emre,M.,McKeith,I.,DeDeyn,P.P.,Aarsland,D.,Hsu,C.,andLane,R.006.Eectsofrivastigmineinpatientswithandwithoutvisualhalluci-nationsindementiaassociatedwithParkinson'sdisease.MovementDisorders,21,1899-1907.Butler,J.C.982.Theclosureproblemasreectedindiscriminantfunctionanalysis.ChemicalGeology,37,367-375.CacabelosR,TakedaM,andWinbladB.99.Theglutamatergicsystemandneurodegenerationindementia:PreventivestrategiesinAlzheimer'sdisease.InternationalJournalofGeriatricPsychiatry,14,3-47.Caeyenberghs,K.,Balschun,D.,Roces,D.P.,Schwake,M.,Saftig,P.,andD'Hooge,R.2006.Multivariateneurocognitiveandemotionalproleofamannosidosismurinemodelfortherapyassessment.NeurobiologyofDisease,23,422-432.Cai,H.,Wang,Y.,McCarthy,D.,Wen,H.,Borchelt,D.R.,Price,D.L.,andWong,P.C.2001.BACE1isthemajor-secretaseforgenerationofApeptidesbyneurons.NatureNeuroscience,4,233-234.Calhoun,M.,Wiederhold,K.,Abramowski,D.,Phinney,A.,Probst,A.,Sturchler-Pierrat,C.,Staufenbiel,M.,Sommer,B.,andJucker,M.998.NeuronlossinAPPtransgenicmice.Nature,395,755-756.Campion,D.,Brice,A.,Hannequin,D.,Tardieu,S.,Dubois,B.,Calenda,A.,Brun,E.,Penet,C.,Tayot,J.,Martinez,M.,Bellis,M.,Mallet,J.,Agid,Y.,andClerget-Darpoux,F.995.Alargepedigreewithearly-onsetAlzheimer'sdisease:clinical,neuropathologic,andgeneticcharacterization.Neurology,45,80-85.Canevari,L.,Abramov,A.Y.,andDuchen,M.R.2004.Toxicityofamyloidbetapeptide:Talesofcalcium,mitochondria,andoxidativestress.NeurochemicalResearch,29,637-650.Carlson,G.A.,Borchelt,D.R.,Dake,A.,Turner,S.,Danielson,V.,Con,J.D.,Eckman,C.,Meiners,J.,Nilsen,S.P.,Younkin,S.G.,andHsiao,K.K.1997.Geneticmodicationofthephenotypesproducedbyamyloidprecursorproteinoverexpressionintransgenicmice.HumanMolecularGenetics,6,1951-1959.Carter,R.J.,Lione,L.A.,Humby,T.,Mangiarini,L.,Mahal,A.,Bates,G.P.,Morton,A.J.,andDunnett,S.B.99.CharacterizationofprogressivemotordecitsinmicetransgenicforthehumanHuntington'sdiseasemutation.JournalofNeuroscience,19,3248-3257.Cattell,R.B.966.Thescreetestforthenumberoffactors.MultivariateBe-havioralResearch,1,245-276.254

PAGE 265

Cerhan,J.H.,Ivnik,R.J.,Smith,G.E.,Tangalos,E.C.,Petersen,R.C.,andBoeve,B.F.02.Diagnosticutilityofletteruency,categoryuency,anduencydierencescoresinAlzheimer'sdisease.ClinicalNeuropsychologist,16,35-42.Chae,J.H.,Jeong,J.,Peterson,B.S.,Kim,D.J.,Bahk,W.M.,Jun,T.Y.,Kim,S.Y.,andKim,K.S.004.DimensionalcomplexityoftheEEGinpatientswithposttraumaticstressdisorder.PsychiatryResearch,131,79-89.Chai,C.K.2007.ThegeneticsofAlzheimer'sdisease.AmericanJournalofAlzheimer'sDiseaseandOtherDementias,22,37-41.Chandrasekaran,B.,andJain,A.K.1979.Onbalancingdecisionfunctions.JournalofCyberneticsandInformationScience,2,12-15.Chapman,P.F.,White,G.L.,Jones,M.W.,Cooper-Blacketer,D.,Marshall,V.J.,Irizarry,M.,Younkin,L.,Good,M.A.,Bliss,T.V.,Mayman,B.T.,Younkin,S.G.,andHsiao,K.K.999.Impairedsynapticplasticityandlearninginagedamyloidprecursorproteintransgenicmice.NatureNeuroscience,2,271-276.Chen,G.,Chen,K.,Knox,J.,Inglis,J.,Bernard,A.,Martin,S.,Justice,A.,McConlogue,L.,Games,D.,Freedman,S.,andMorris,R.2000.Alearningdecitrelatedtoageandbeta-amyloidplaquesinamousemodelofAlzheimer'sdisease.Nature,408,975-979.Cheng,D.H.,andTang,X.C.98.ComparativestudiesofhuperzineA,E2020,andtacrineonbehaviorandcholinesteraseactivities.Pharmacology,Biochem-istry,andBehavior,60,377-386.Cho,Y.H.,Delcasso,S.,Israel,A.,andJeantet,Y.007.Alonglistvisuo-spatialsequentiallearninginmice.BehavioralBrainResearch,179,152-158.Chou,T.992.Wakeupandsmellthecoee:Caeine,coee,andthemedicalconsequences.WesternJournalofMedicine,157,544-553.Christen,Y.2000.OxidativestressandAlzheimer'sdisease.AmericanJournalofClinicalMedicine,71,621S-629S.Christie,M.A.,andDalrymple-Alford,J.C.04.Anewratmodelofthehu-manserialreactiontimetask:Contrastingeectsofcaudateandhippocampallesions.JournalofNeuroscience,24,1034-1039.Christie,M.A.,andHersch,S.M.004.Demonstrationofnondeclarativese-quencelearninginmice:Developmentofananimalanalogofthehumanserialreactiontimetask.LearningandMemory,11,720-723.255

PAGE 266

Citron,M.,Diehl,T.S.,Gordon,G.,Biere,A.L.,Seubert,P.,andSelkoe,D.J.1996.Evidencethatthe42-and40-aminoacidformsofamyloidbetaproteinaregeneratedfromthebeta-amyloidprecursorproteinbydierentproteaseactivities.ProceedingsoftheNationalAcademyofSciencesUSA,93,13170-13175.Cohen,F.E.99.Proteinmisfoldingandpriondiseases.JournalofMolecularBiology,293,313-320.Cohen,F.E.,andKelly,J.W.2003.Therapeuticapproachestoprotein-misfoldingdiseases.Nature,426,905-909.Cohen,I.L.,Sudhalter,V.,Landon-Jimenez,D.,andKeogh,M.93.Aneu-ralnetworkapproachtotheclassicationofautism.JournalofAutismandDevelopmentalDisorders,23,443-466.Cohen,J.60.Acoecientofagreementfornominalscales.EducationalandPsychologicalMeasurement,20,37-46.Cohen,J.D.,Dunbar,K.,andMcClelland,J.M.990.Onthecontrolofauto-maticprocesses:AparalleldistributedprocessingaccountoftheStroopeect.PsychologicalReview,97,332-361.Cohen,J.D.,Perlstein,W.M.,Braver,T.S.,Nystrom,L.E.,Noll,D.C.,Jonides,J.,andSmith,E.E.997.Temporaldynamicsofbrainactivationduringaworkingmemorytask.Nature,386,604-608.Cohen,J.S.,Sturdy,C.,andHicks,M.96.Intratrialproactiveinterferenceinrats'serialalternationperformanceintheradialmaze.AnimalLearningandBehavior,24,300-309.Cohen,N.J.,Eichenbaum,H.,Deacedo,B.S.,andCorkin,S.85.Dierentmemorysystemsunderlyingacquisitionofproceduralanddeclarativeknowl-edge.AnnalsoftheNewYorkAcademyofSciences,444,54-71.Cohen-Manseld,J.2001.Nonpharmacologicinterventionsforinappropriatebe-haviorsindementia:Areview,summary,andcritique.AmericanJournalofGeriatricPsychiatry,9,361-381.Cohen-Salmon,C.,Venault,P.,Martin,B.,Raalli-Sebille,M.J.,Barkats,M.,Clostre,F.,Pardon,M.C.,Christen,Y.,andChapouthier,G.997.EectsofGinkgobilobaextractEGb761onlearningandpossibleactionsonaging.JournalofPhysiology,91,291-300.Cole,M.R.,andChappell-Stephenson,R.03.Exploringthelimitsofspatialmemoryinrats,usingverylargemazes.LearningandBehavior,31,349-368.256

PAGE 267

Come,J.H.,Fraser,P.E.,andLansbury,P.T.,Jr.1993.Akineticmodelforamyloidformationinthepriondiseases:Importanceofseeding.ProceedingsoftheNationalAcademyofSciencesUSA,90,5959-5963.Conover,W.J.,andIman,R.L.81.Ranktransformationsasabridgebetweenparametricandnonparametricstatistics.AmericanStatistician,35,124-129.Corder,E.H.,Saunders,A.M.,Strittmatter,W.J.,Schmechel,D.E.,Gaskell,P.C.,Small,G.W.,Roses,A.D.,Haines,J.L.,andPericak-Vance,M.A.1993.GenedoseofapolipoproteinEtype4alleleandtheriskofAlzheimer'sdiseaseinlateonsetfamilies.Science,261,921-923.Cortes,C.,andVapnik,V.1995.Supportvectornetworks.MachineLearning,20,273-297.Costa,D.A.,Cracchiolo,J.R.,Bachstetter,A.D.,Hughes,T.F.,Bales,K.R.,Paul,S.M.,Mervis,R.F.,Arendash,G.W.,andPotter,H.007.En-richmentimprovescognitioninADmicebyamyloid-relatedandunrelatedmechanisms.NeurobiologyofAging,28,831-844.Cotman,C.W.,Head,E.,Muggenburg,B.A.,Zicker,S.,andMilgram,N.W.002.Brainaginginthecanine:Adietenrichedinantioxidantsreducescognitivedysfunction.NeurobiologyofAging,23,809-818.Cox,P.A.,andBalick,M.J.994.Theethnobotanicalapproachtodrugdiscov-ery.ScienticAmerican,2706,82-87.Crabbe,J.C.,andMorris,R.G.M.004.Festinalente:Late-nightthoughtsonhigh-throughputscreeningofmousebehavior.NatureNeuroscience,7,1175-1179.Cracchiolo,J.R.,Mori,T.,Nazian,S.J.,Tan,J.,Potter,H.,andArendash,G.W.007.Enhancedcognitiveactivity{overandabovesocialorphysicalactivity{isrequiredtoprotectAlzheimer'smiceagainstcognitiveimpairment,reduceAdeposition,andincreasesynapticimmunoreactivity.NeurobiologyofLearningandMemory,88,277-294.Crawley,J.N.999.Behavioralphenotypingoftransgenicandknockoutmice:Experimentaldesignandevaluationofgeneralhealth,sensoryfunctions,motorabilities,andspecicbehavioraltests.BrainResearch,835,18-26.Crawley,J.N.,Belknap,J.K.,Collins,A.,Crabbe,J.C.,Frankel,W.,Henderson,N.,Hitzemann,R.J.,Maxson,S.C.,Miner,L.L.,Silva,A.J.,Wehner,J.M.,Wynshaw-Boris,A.,andPaylor,R.997.Behavioralphenotypesofinbredmousestrains:Implicationsandrecommendationsformolecularstudies.Psychopharmacology,132,107-124.257

PAGE 268

Crisler,S.,Morrissey,M.J.,Anch,A.M.,andBarnett,D.W.008.Sleep-stagescoringintheratusingasupportvectormachine.JournalofNeuroscienceMethods,168,524-534.Cristianini,N.,andShawe-Taylor,J.000.AnIntroductiontoSupportVectorMachines.CambridgeUniversityPress.Crum,R.M.,Anthony,J.C.,Bassett,S.S.,andFolstein,M.F.93.Population-basednormsfortheMiniMentalStateExaminationbyageandeducationallevel.JournaloftheAmericanMedicalAssociation,269,2386-2391.Crusio,W.E.,andSchwegler,H.005.Learningspatialorientationtasksintheradial-mazeandstructuralvariationinthehippocampusininbredmice.BehavioralandBrainFunctions,1,3.Crusio,W.E.,Schwegler,H.,andBrust,I.995.Covariationsbetweenhippocam-palmossybersandworkingandreferencememoryinspatialandnon-spatialradialmazetasksinmice.EuropeanJournalofNeuroscience,5,1413-1420.Cummings,B.J.,Head,E.,Afagh,A.J.,Milgram,N.W.,andCotman,C.W.996a.Beta-amyloidaccumulationcorrelateswithcognitivedysfunctionintheagedcanine.NeurobiologyofLearningandMemory,66,11-23.Cummings,B.J.,Head,E.,Ruehl,W.,Milgram,N.W.,andCotman,C.W.996b.Thecanineasananimalmodelofhumanaginganddementia.Neu-robiologyofAging,17,259-268.Cummings,B.J.,Satou,T.,Head,E.,Milgram,N.W.,Cole,G.M.,Savage,M.J.,Podlisny,M.B.,Selkoe,D.J.,Siman,R.,Greenberg,B.D.,andCotman,C.W.96c.DiuseplaquescontainC-terminalA42andnotA40:Evidencefromcatsanddogs.NeurobiologyofAging,17,653-659.Cummings,J.L.00.CognitiveandbehavioralheterogeneityinAlzheimer'sdisease:Seekingtheneurobiologicalbasis.NeurobiologyofAging,21,845-861.Cummings,J.L.,andZhong,K.2006.Treatmentsforbehavioraldisordersinneurodegenerativediseases:Drugdevelopmentstrategies.NatureReviews,5,64-74.Cunningham,P.,Carney,J.,andJacob,S.2000.Stabilityproblemswitharticialneuralnetworksandtheensemblesolution.ArticialIntelligenceinMedicine,20,217-225.Cupp,C.J.,Jean-Philippe,C.,Kerr,W.W.,Patil,A.R.,andPerez-Camargo,G.006.Eectofnutritionalinterventionsonlongevityofseniorcats.Interna-tionalJournalofAppliedResearchinVeterinaryMedicine,4,34-50.258

PAGE 269

Cybenko,G.989.Approximationbysuperpositionsofasigmoidalfunction.MathematicsofControl,Signals,andSystems,2,303-314.Dall'Igna,O.P.,Fett,P.,Gomes,M.W.,Souza,D.O.,Cunha,R.A.,andLara,D.R.007.CaeineandadenosineA2areceptorantagonistsprevent-amyloid5-35-inducedcognitivedecitsinmice.ExperimentalNeurology,203,241-245.DaSilva,K.A.,Brown,M.E.,Westaway,D.,andMcLaurin,J.06.Immu-nizationwithamyloid-usingGM-CSFandIL-4reducesamyloidburdenandaltersplaquemorphology.NeurobiologyofDisease,23,433-444.DeCoteau,W.E.,andKesner,R.P.00.Adoubledissociationbetweentherathippocampusandmedialcaudo-putameninprocessingtwoformsofknowledge.BehavioralNeuroscience,114,1096-1108.DeFigueiredo,R.J.P.,Shankle,W.R.,Maccato,A.,Dick,M.B.,Mundkur,P.,Mena,I.,andCotman,C.W.95.Neural-network-basedclassicationofcognitivelynormal,demented,Alzheimerdiseaseandvasculardementiafromsinglephotonemissionwithcomputedtomographyimagedatafrombrain.ProceedingsoftheNationalAcademyofSciencesUSA,92,5530-5534.DeKosky,S.,Fitzpatrick,A.,Ives,D.,Saxton,J.,Williamson,J.,Lopez,O.,Burke,G.,Fried,L.,Kuller,L.,Robbins,J.,Tracy,R.,Woolard,N.,Dunn,L.,Kron-mal,R.,Nahin,R.,andFurberg,C.006.TheGinkgoEvaluationofMemoryGEMstudy:DesignandbaselinedataofarandomizedtrialofGinkgobilobaextractinpreventionofdementia.ContemporaryClinicalTrials,27,238-253.Del-Favero,J.,Goossens,D.,VandenBossche,D.,andVanBroeckhoven,C.999.YACfragmentationwithrepetitiveandsingle-copysequences:detailedphysicalmappingofthepresenilin1geneonchromosome14.Gene,229,193-201.DeMager,P.P.,Penke,B.,Walter,R.,Harkany,T.,andHartignny,W.002.PathologicalpeptidefoldinginAlzheimer'sdiseaseandotherconformationaldisorders.CurrentMedicinalChemistry,9,1763-1780.DenDunnen,W.F.A.,Brouwer,W.H.,Bijlard,E.,Kamphuis,J.,vanLinschoten,K.,Eggens-Meijer,E.,andHolstege,G.008.Nodiseaseinthebrainofa115-year-oldwoman.NeurobiologyofAging,29,1127-1132.DeQuervain,D.J.,Henke,K.,Aerni,A.,Coluccia,D.,Wollmer,M.A.,Hock,C.,Nitsch,R.M.,andPapassotiropoulos,A.2003.Afunctionalgeneticvariationofthe5-HT2areceptoraectshumanmemory.NatureNeuroscience,6,1141-1142.259

PAGE 270

DeQuervain,D.J.,andPapassotiropoulos,A.2006.Identicationofageneticclusterinuencingmemoryperformanceandhippocampalactivityinhumans.ProceedingsoftheNationalAcademyofSciencesUSA,103,4270-4274.Dere,E.,Huston,J.P.,DeSouzaSilva,M.A.005.Episodic-likememoryinmice:Simultaneousassessmentofobject,placeandtemporalordermemory.BrainResearchProtocols,16,10-19.Devanand,D.P.,Jacobs,D.M.,Tang,M.X.,DelCastillo-Castenada,C.,Sano,M.,Marder,K.,Bell,K.,Bylsma,F.W.,Brandt,J.,Albert,M.,andStern,Y.997.ThecourseofpsychopathologicfeaturesinmildtomoderateAlzheimer'sdisease.ArchivesofGeneralPsychiatry,54,257-263.D'Hooge,R.,andDeDeyn,P.P.2001.ApplicationsoftheMorriswatermazeinthestudyoflearningandmemory.BrainResearchReviews,36,60-90.Dierckx,E.,Engelborghs,S.,DeRaedt,R.,DeDeyn,P.P.,andPonjaert-Kristoersen,I.07.Dierentiationbetweenmildcognitiveimpairment,Alzheimer'sdis-easeanddepressionbymeansofcuedrecall.PsychologicalMedicine,37,747-755.Dietterich,T.97.Machinelearningresearch:Fourcurrentdirections.AIMagazine,18,97-136.Dodart,J.-C.,Mathis,C.,Bales,K.R.,andPaul,S.M.02.DoesmymousehaveAlzheimer'sdisease?Genes,BrainandBehavior,1,142-155.Dodart,J.,Mathis,C.,Saura,J.,Bales,K.,Paul,S.,andUngerer,A.2000.Neu-roanatomicalabnormalitiesinbehaviorallycharacterizedAPP[V717F]trans-genicmice.NeurobiologyofDisease,7,71-85.Dodart,J.,Meziane,H.,Mathis,C,Ungerer,A.,Bales,K.,andPaul,S.999.BehavioraldisturbancesintransgenicmiceoverexpressingtheV717Fbeta-amyloidprecursorprotein.BehavioralNeuroscience,113,982-990.Dolinoy,D.C.,Weidman,J.R.,andJirtle,R.L.2007.Epigeneticgeneregula-tion:Linkingearlydevelopmentalenvironmenttoadultdisease.ReproductiveToxicology,23,297-307.Donovan,M.H.,Yazdani,U.,Norris,R.D.,Games,D.,German,D.C.,andEisch,A.J.006.DecreasedadulthippocampalneurogenesisinthePDAPPmousemodelofAlzheimer'sdisease.JournalofComparativeNeurology,495,70-83.Dougall,N.J.,Bruggink,S.,andEbmeier,K.P.004.Systematicreviewofthediagnosticaccuracyof99mTc-HMPAO-SPECTindementia.AmericanJournalofGeriatricPsychiatry,12,554-570.260

PAGE 271

Doyon,B.992.Ontheexistenceandroleofchaoticprocessesinthenervoussystem.ActaBiotheoretica,40,113-119.Dubois,B.,Feldman,H.H.,Jacova,C.,Dekosky,S.T.,Barberger-Gateau,P.,Cummings,J.,Delacourte,A.,Galasko,D.,Gauthier,S.,Jicha,G.,Meguro,K.,O'Brien,J.,Pasquier,F.,Robert,P.,Rossor,M.,Salloway,S.,Stern,Y.,Visser,P.J.,andScheltens,P.2007.ResearchcriteriaforthediagnosisofAlzheimer'sdisease:RevisingtheNINCDS-ADRDAcriteria.LancetNeurol-ogy,6,734-746.Duin,R.P.W.96.Anoteoncomparingclassiers.PatternRecognitionLetters,17,529-536.Dunkin,J.J.,Leuchter,A.F.,Newton,T.F.,andCook,I.A.94.ReducedEEGcoherenceindementia:Stateortraitmarker?BiologicalPsychiatry,35,870-879.Ebert,U.,Grossmann,M.,Oertel,R.,Gramatte,T.,andKirch,W.01.Pharma-cokinetic-pharmacodynamicmodelingoftheelectroencephalogrameectsofscopolamineinhealthyvolunteers.JournalofClinicalPharmacology,41,51-60.Ebert,U.,andKirch,W.98.Scopolaminemodelofdementia:Electroen-cephalogramndingsandcognitiveperformance.EuropeanJournalofClinicalInvestigation,28,944-949.Eckelinen,M.,Ngandu,T.,Tuomilehto,J.,Soininen,H.,andKivipelto,M.008.Midlifecoeeandteadrinkingandtheriskoflate-lifedementia:Apopulation-basedCAIDEstudy.JournalofAlzheimer'sDisease,16,xxx-xxx.Efron,B.83.Estimatingtheerrorrateofapredictionrule:Improvementoncross-validation.JournaloftheAmericanStatisticalAssociation,78,316-331.Egan,M.F.,Kojima,M.,Callicott,J.H.,Goldberg,T.E.,Kolachana,B.S.,Bertolino,A.,Zaitsev,E.,Gold,B.,Goldman,D.,Dean,M.,Lu,B.,andWeinberger,D.R.003.TheBDNFVal66Metpolymorphismaectsactivity-dependentsecretionofBDNFandhumanmemoryandhippocampalfunction.Cell,112,257-269.Eichenbaum,H.1999.Thehippocampusandmechanismsofdeclarativememory.BehavioralBrainResearch,103,123-133.Eichenbaum,H.,Stewart,C.,andMorris,R.G.M.90.Hippocampalrepresen-tationinplacelearning.JournalofNeuroscience,10,3531-3542.261

PAGE 272

El-Naqa,I.,Yongyi,Y.,Wernick,M.N.,Galatsanos,N.P.,andNishikawa,R.M.002.Asupportvectormachineapproachfordetectionofmicrocalcications.IEEETransactionsonMedicalImaging,21,1552-1563.Emilien,G.,Maloteaux,J.-M.,Beyreuther,K.,andMasters,C.L.2000.Alzheimerdisease:Mousemodelspavethewayfortherapeuticopportunities.ArchivesofNeurology,57,176-181.Emre,M.,Aarsland,D.,Albanese,A.,Byrne,E.J.,Deuschl,G.,DeDeyn,P.P.,Durif,F.,Kulisevsky,J.,vanLaar,T.,Lees,A.,Poewe,W.,Robillard,A.,Rosa,M.M.,Wolters,E.,Ouarg,P.,Tekin,S.,andLane,R.2004.RivastigminefordementiaassociatedwithParkinson'sdisease.NewEnglandJournalofMedicine,351,2509-2518.Eriksen,J.L.,Sagi,S.A.,Smith,T.E.,Weggen,S.,Das,P.,McLendon,D.C.,Ozols,V.V.,Jessing,K.W.,Zavitz,K.H.,Koo,E.H.,andGolde,T.E.003.NSAIDsandenantiomersofurbiprofentargetgamma-secretaseandlowerAbeta42invivo.JournalofClinicalInvestigation,112,440-449.Ethell,D.W.,Shippy,D.,Cao,C.,Cracchiolo,J.R.,Runfeldt,M.,Blake,B.,andArendash,G.W.006.A-specicT-cellsreversecognitivedeclineandsynapticlossinAlzheimer'smice.NeurobiologyofDisease,23,351-361.Evans,D.A.,Funkenstein,H.H.,Albert,M.S.,Scherr,P.A.,Cook,N.R.,Chown,M.J.,Hebert,L.E.,Hennekens,C.H.,andTaylor,J.O.89.PrevalenceofAlzheimer'sdiseaseinacommunitypopulationofolderpersons.Higherthanpreviouslyreported.JournaloftheAmericanMedicalAssociation,262,2551-2556.Fabrigar,L.R.,Wegener,D.T.,MacCallum,R.C.,andStrahan,E.J.999.Evaluatingtheuseofexploratoryfactoranalysisinpsychologicalresearch.PsychologicalMethods,3,272-299.Farias,S.T.,Harrell,E.,Neumann,C.,andHoutz,A.2003.Therelationshipbe-tweenneuropsychologicalperformanceanddailyfunctioninginindividualswithAlzheimer'sdisease:Ecologicalvalidityofneuropsychologicaltests.ArchivesofClinicalNeuropsychology,18,655-672.Farina,E.,Fioravanti,R.,Chiavari,L.,Imbornone,E.,Alberoni,M.,Pomanti,S.,Pinardi,G.,Pignatti,R.,andMariani,C.2002.ComparingtwoprogramsofcognitivetraininginAlzheimer'sdisease:Apilotstudy.ActaNeurologicaScandinavica,105,365-371.Farlow,M.,Gracon,S.I.,Hershey,L.A.,Lewis,K.W.,Sadowsky,C.H.,Dolan-Ureno,J.992.AcontrolledtrialoftacrineinAlzheimer'sdisease.JournaloftheAmericanMedicalAssociation,268,2523-2529.262

PAGE 273

Fawcett,T.2006.AnintroductiontoROCanalysis.PatternRecognitionLetters,27,861-874.Fayyad,U.,Piatetsky-Shapiro,G.,andSmyth,P.1996a.Fromdataminingtoknowledgediscoveryindatabases.AIMagazine,Fall,37-54.Fayyad,U.,Piatetsky-Shapiro,G.,andSmyth,P.96b.TheKDDprocessforextractingusefulknowledgefromvolumesofdata.CommunicationsoftheACM,391,27-34.Feldmann,H.,Gauthier,S.,Hecker,J.,Vellas,B.,Subbiah,P.andWhalen,E.001.A24-week,randomized,double-blindstudyofdonepezilinmoderatetosevereAlzheimer'sdisease.Neurology,57,613-620.Fell,J.,Fernandez,G.,andElger,C.E.03.Morethansynchrony:EEGchaotic-itymaybenecessaryforconsciousbrainfunctioning.MedicalHypotheses,61,158-160.Fernandes,C.,Gonzalez,M.I.,Wilson,C.A.,andFile,S.E.999.Factoranalysisshowsthatfemaleratbehaviorischaracterizedprimarilybyactivity,maleratsaredrivenbysexandanxiety.Pharmacology,Biochemistry,andBehavior,64,731-738.Fillenbaum,G.G.,Heyman,A.,Wilkinson,W.E.,andHaynes,C.S.987.Com-parisonoftwoscreeningtestsinAlzheimer'sdisease.ArchivesofNeurology,44,924-927.Fisher,N.J.,Rourke,B.P.,andBieliauskas,L.A.99.NeuropsychologicalsubgroupsofpatientswithAlzheimer'sdisease:Anexaminationoftherst10yearsofCERADdata.JournalofClinicalandExperimentalNeuropsychology,21,488-518.Fisher,R.A.936.Theuseofmultiplemeasurementsintaxonomicproblems.AnnalsofEugenics,7,179-188.Fleiss,J.L.971.Measuringnominalscaleagreementamongmanyraters.Psy-chologicalBulletin,76,378-382.Fleminger,S.,Oliver,D.L.,Lovestone,S.,Rabe-Hesketh,S.,andGiora,A.003.HeadinjuryasariskfactorforAlzheimer'sdisease:Theevidence10yearson.Apartialreplication.JournalofNeurology,NeurosurgeryandPsychiatry,74,857-862.Flint,J.99.Thegeneticbasisofcognition.Brain,122,2015-2031.263

PAGE 274

Folstein,M.F.,Folstein,S.F.,andMcHugh,P.R.75.Mini-MentalStateExam:Apracticalmethodforgradingthecognitivestateofpatientsfortheclinician.JournalofPsychiatricResearch,12,189-198.Forbes,K.E.,Shanks,M.F.,andVenneri,A.2004.TheevolutionofdysgraphiainAlzheimer'sdisease.BrainResearchBulletin,63,19-24.Forstl,H.,andKurtz,A.1999.ClinicalfeaturesofAlzheimer'sdisease.EuropeanArchivesofPsychiatryandClinicalNeurosciences,249,288-290.Foster,D.J.,Morris,R.G.M.,andDayan,P.000.Amodelofhippocampallydependentnavigation,usingthetemporaldierencelearningrule.Hippocam-pus,10,1-16.Fox,K.R.,Stathi,A.,McKenna,J.,andDavis,M.G.07.Physicalactivityandmentalwell-beinginolderpeopleparticipatingintheBetterAgeingProject.EuropeanJournalofAppliedPhysiology,100,591-602.Fraga,M.F.,Ballestar,E.,Paz,M.F.,Ropero,S.,Setien,F.,Ballestar,M.L.,Heine-Suner,D.,Cigudosa,J.C.,Urioste,M.,Benitez,J.,Boix-Chornet,M.,Sanchez-Aguilera,A.,Ling,C.,Carlsson,E.,Poulsen,P.,Vaag,A.,Stephan,Z.,Spector,T.D.,Wu,Y.-Z.,Plass,C.,andEsteller,M.2005.Epigeneticdierencesariseduringthelifetimeofmonozygotictwins.ProceedingsoftheNationalAcademyofSciencesUSA,102,10604-10609.Fratiglioni,L.,Paillard-Borg,S.,andWinblad,B.04.Anactiveandsociallyin-tegratedlifestyleinlatelifemightprotectagainstdementia.LancetNeurology,3,343-353.Fredholm,B.,Battig,K.,Holmen,J.,Nehlig,A.,andZvartau,E.99.Actionsofcaeineinthebrainwithspecialreferencetofactorsthatcontributetoitswidespreaduse.PharmacologicalReviews,51,83-133.French,B.M.,Dawson,M.R.W.,andDobbs,A.R.97.Classicationandstag-ingofdementiaoftheAlzheimertype:Acomparisonbetweenneuralnetworksandlineardiscriminantanalysis.ArchivesofNeurology,54,1001-1009.Frohlich,H.,Hoenselaar,A.,Eichner,J.,Rosenbrock,H.,Birk,G.,andZell,A.008.Automatedclassicationofthebehaviorofratsintheforcedswimmingtestwithsupportvectormachines.NeuralNetworks,21,92-101.Fuld,P.A.81.FuldObject-MemoryEvaluation.WoodDale,IL:StoeltingCo.Gaasterland,T.,andBekiranov,S.2000.Makingthemostofmicroarraydata.NatureGenetics,24,204-206.264

PAGE 275

Gaito,J.958.ThesingleLatinsquaredesigninpsychologicalresearch.Psy-chometrika,23,369-378.Gaito,J.1961.Repeatedmeasurementdesignsandcounterbalancing.Psycholog-icalBulletin,58,46-54.Gallagher,M.,andRapp,P.R.997.Theuseofanimalmodelstostudytheeectsofagingoncognition.AnnualsReviewsinPsychology,48,339-370.Gallez,D.,andBabloyantz,A.991.PredictabilityofhumanEEG:Adynamicalapproach.BiologicalCybernetics,64,381-391.Gallistel,C.R.,Fairhurst,S.,andBalsam,P.04.Thelearningcurve:Im-plicationsofaquantitativeanalysis.ProceedingsoftheNationalAcademyofSciencesUSA,101,13124-13131.Galsworthy,M.,Paya-Cano,J.,Liu,L.,Monleon,S.,Gregoryan,G.,Fernandes,C.,Schalkwyk,L.,andPlomin,R.2005.Assessingreliability,heritabilityandgeneralcognitiveabilityinabatteryofcognitivetasksforlaboratorymice.BehaviorGenetics,35,675-692.Games,D.,Adams,D.,Alessandrini,R.,Barbour,R.,Berthelette,P.,Blackwell,C.,Carr,T.,Clemens,J.,Donaldson,T.,Gillespie,F.,Guido,T.,Hagopian,S.,Johnson-Wood,K.,Khan,K.,Lee,M.,Leibowitz,P.,Lieberburg,I.,Little,S.,Masliah,E.,McConlogue,L.,Montoya-Zavala,M.,Mucke,L.,Paganini,L.,Penniman,E.,Power,M.,Schenk,D.,Seubert,P.,Snyder,B.,Soriano,F.,Tan,H.,Vitale,J.,Wadsworth,S.,Wolozin,B.,andZhao,J.995.Alzheimer-typeneuropathologyintransgenicmiceoverexpressingV717Fbeta-amyloidprecursorprotein.Nature,373,523-527.Gamst,G.,Meyers,L.S.,andGuardino,A.J.008.AnalysisofVarianceDesigns:AConceptualandComputationalApproachwithSPSSandSAS.NewYork:CambridgeUniversityPress.Pp.420-450.Ganguli,M.,Chandra,V.,Kamboh,M.I.,Johnston,J.M.,Dodge,H.H.,Thelma,B.K.,Juyal,R.C.,Pandav,R.,Belle,S.H.,andDeKosky,S.T.2000.ApolipoproteinEpolymorphismandAlzheimerdisease:TheIndo-USCross-NationalDementiaStudy.ArchivesofNeurology,57,824-830.Garcia-Perez,E.,Violante,A.,andCervantes-Perez,F.1998.Usingneuralnet-worksfordierentialdiagnosisofAlzheimerdiseaseandvasculardementia.ExpertSystemswithApplications,14,219-225.Garrard,P.,Maloney,L.M.,Hodges,J.R.,andPatterson,K.005.TheeectsofveryearlyAlzheimer'sdiseaseonthecharacteristicsofwritingbyarenownedauthor.Brain,128,250-260.265

PAGE 276

Gauthier,S.2001.Alzheimer'sdisease:Currentandfuturetherapeuticperspec-tives.ProgressinNeuropsychopharmacologyandBiologicalPsychiatry,25,73-89.Gauthier,S.,Vellas,B.,Farlow,M.,andBurn,D.006.Anaggressivecourseofdiseaseindementia.Alzheimer'sandDementia,2,210217.Gearing,M.,Rebeck,G.W.,Hyman,B.T.,Tigges,J.,andMirra,S.S.994.NeuropathologyandapolipoproteinEproleofagedchimpanzees:Implica-tionsforAlzheimerdisease.ProceedingsoftheNationalAcademyofSciencesUSA,91,9382-9386.Gearing,M.,Tigges,J.,Mori,H.,andMirra,S.S.97.-amyloidAdeposi-tioninthebrainsofagedorangutans.NeurobiologyofAging,18,139-146.Gee,J.,Ding,L.,Xie,Z.,Lin,M.,DeVita,C.andGrossman,M.2003.Alzheimer'sdiseaseandfrontotemporaldementiaexhibitdistinctatrophy-behaviorcorre-lates:acomputer-assistedimagingstudy.AcademicRadiology,10,1392-1409.Gehan,E.1965.AgeneralizedWilcoxontestforcomparingarbitrarilysingly-censoredsamples.Biometrika,52,203-223.Gerlai,R.1996.Gene-targetingstudiesofmammalianbehavior:Isitthemutationorthebackgroundgenotype?TrendsinNeuroscience,19,177-186.Gevins,A.,Smith,M.E.,Leong,H.,McEvoy,L.,Whiteld,S.,Du,R.,andRush,G.98.Monitoringworkingmemoryloadduringcomputer-basedtaskswithEEGpatternrecognitionmethods.HumanFactors,40,79-91.Gevins,A.,Smith,M.E.,McEvoy,L.,andYu,D.97.HighresolutionEEGmappingofcorticalactivationrelatedtoworkingmemory:Eectsoftaskdif-culty,typeofprocessing,andpractice.CerebralCortex,7,374-385.Giaquinto,S.,andNolfe,G.86.TheEEGinthenormalelderly:Acontribu-tiontotheinterpretationofaginganddementia.ElectroencephalographyandClinicalNeurophysiology,63,540-546.Giarratano,J.C.,andRiley,G.D.04.ExpertSystems:PrinciplesandPro-gramming,4thEd.Florence,KY:CourseTechnology.Gimenez-Llort,L.,Blazquez,G.,Canete,T.,Johansson,B.,Oddo,S.,Tobena,A.,LaFerla,F.M.,andFernandez-Teruel,A.007.ModelingbehavioralandneuronalsymptomsofAlzheimer'sdiseaseinmice:Aroleforintraneuronalamyloid.NeuroscienceandBiobehavioralReviews,31,125-147.Glass,L.001.Synchronizationandrhythmicprocessesinphysiology.Nature,410,277-284.266

PAGE 277

Glenner,G.G.,andWong,C.W.984a.Alzheimer'sdisease:Initialreportofthepuricationandcharacterizationofanovelcerebrovascularamyloidprotein.BiochemicalandBiophysicalResearchCommunications,120,885-890.Glenner,G.G.,andWong,C.W.984b.Alzheimer'sdiseaseandDown'ssyn-drome:Sharingofaauniquecerebrovascularamyloidbrilprotein.Biochem-icalandBiophysicalResearchCommunications,122,1131-1135.Godkar,P.B.,Gordon,R.K.,Ravindran,A.,andDoctor,B.P.004.Celastruspaniculatusseedwatersolubleextractsprotectagainstglutamatetoxicityinneuronalculturesfromratforebrain.JournalofEthnopharmacology,93,213-219.Goldgaber,D.,Lerman,M.I.,McBride,O.W.,Saotti,U.,andGajdusek,D.C.987.CharacterizationandchromosomallocalizationofacDNAencodingbrainamyloidofAlzheimer'sdisease.Science,235,877-880.Golub,T.,Slonim,D.,Tamayo,P.,Huard,C.,Gaansenbeek,M.,Mesirov,J.H.H.C.,Loh,M.,Downing,J.,Caligiuri,M.,Bloomeld,C.,andLander,E.1999.Molecularclassicationofcancer:Classdiscoveryandclasspredictionbygeneexpressionmonitoring.Science,286,531-537.Gomez,R.G.,andWhite,D.A.06.UsingverbaluencytodetectverymilddementiaoftheAlzheimertype.ArchivesofClinicalNeuropsychology,21,771-775.Gonzalez,F.J.2002.Theperoxisomeproliferator-activatedreceptorPPAR:Roleinhepatocarcinogenesis.MolecularandCellularEndocrinology,193,71-79.Good,M.A.,Hale,G.,andStaal,V.007.Impairedepisodic-like"objectmem-oryinadultAPPswetransgenicmice.BehavioralNeuroscience,121,443-448.Gordon,M.N.,Holcomb,L.A.,Jantzen,P.T.,DiCarlo,G.,Wilcock,D.,Boyett,K.W.,Connor,K.,Melachrino,J.,OCallaghan,J.P.,andMorgan,D.2002.TimecourseofthedevelopmentofAlzheimer-likepathologyinthedoublytransgenicPS1+APPmouse.ExperimentalNeurology,173,183-195.Gordon,M,King,D.,Diamond,D.,Jantzen,P.,Boyett,K.,Hope,C.,Hatcher,J.,DiCarlo,G.,Gottschalk,P.,Morgan,D.andArendash,G.001.Correla-tionbetweencognitivedecitsandAbdepositsintransgenicAPP+PS1mice.NeurobiologyofAging,22,377-386.Goutte,C.997.Noteonfreelunchesandcross-validation.NeuralComputation,9,1211-1215.267

PAGE 278

Graeber,M.B.99.Nomanalone:TherediscoveryofAloisAlzheimer'soriginalcases.BrainPathology,9,237-240.Graeber,M.B.,Kosel,S.,Grasbon-Frodi,F.,Moller,H.J.,andMahraein,P.998.HistopathologyandApoEgenotypeoftherstAlzheimerdiseasepa-tient,AugusteD.Neurogenetics,1,223-228.Graf,P.,andSchacter,D.L.985.Implicitandexplicitmemoryfornewasso-ciationsinnormalandamnesicsubjects.JournalofExperimentalPsychology:Learning,Memory,andCognition,11,501-518.Graf,P.,andSchacter,D.L.87.Selectiveeectsofinterferenceonimplicitandexplicitmemoryfornewassociations.JournalofExperimentalPsychology:Learning,Memory,andCognition,13,45-53.Grassberger,P.,andProcaccia,I.983.Measuringthestrangenessofstrangeattractors.PhysicaD,9,183-208.Graziano,A.,Petrosini,L.,andBartoletti,A.003.Automaticrecognitionofex-ploratorystrategiesintheMorriswatermaze.JournalofNeuroscienceMeth-ods,130,33-44.Green,R.C.,Schneider,L.S.,Zavitz,K.H.,Amato,D.A.,Beelen,A.P.,Swabb,E.A.,andtheTarenurbilPhase3StudyGroup.2008.SafetyandecacyoftarenurbilinsubjectswithmildAlzheimer'sdisease:Resultsfroman18-monthmulti-centerphase3trial.Presentation.Alzheimer'sAssociationInternationalConferenceonAlzheimer'sDisease,July26-31,Chicago,IL.Grice,J.W.2001.Acomparisonoffactorscoresunderconditionsoffactorobliquity.PsychologicalBulletin,6,67-83.Grith,J.S.967.Self-replicationandscrapie.Nature,215,1043-1044.Gron,G.,andRiepe,M.W.04.NeuralbasisforthecognitivecontinuuminepisodicmemoryfromhealthtoAlzheimer'sdisease.AmericanJournalofGeriatricPsychiatry,12,648-652.Grossi,E.,Buscema,M.P.,Snowdon,D.,andAntuono,P.007.Neuropatho-logicalndingsprocessedbyarticialneuralnetworksANNscanperfectlydistinguishAlzheimer'spatientsfromcontrolsintheNunStudy.BioMedCen-tralNeurology,7,15.Grundke-Iqbal,I.,Iqbal,K.,Tung,Y.C.,Quinlan,M.,Wisniewski,H.M.,andBinder,L.I.986.Abnormalphophorylationofthemicrotubule-associatedproteintauinAlzheimercytoskeletalpathology.ProceedingsoftheNationalAcademyofSciencesUSA,83,4913-4917.268

PAGE 279

Guadagnoli,E.,andVelicer,W.F.988.Relationofsamplesizetothestabilityofcomponentpatterns.PsychologicalBulletin,103,265-275.Gunn-Moore,D.,McVee,J.,Bradshaw,J.M.,Pearson,G.R.,Head,E.,andGunn-Moore,F.J.006.Ageingchangesincatbrainsdemonstratedbybeta-amyloidandAT8-immunoreactivephosphorylatedtaudeposits.JournalofFelineMedicineandSurgery,8,234-242.GurejeO.,OgunniyiA.,BaiyewuO.,PriceB.,UnverzagtF.W.,EvansR.M.,Smith-GambleV.,LaneK.A.,GaoS,HallK.S.,HendrieH.C.,andMurrellJ.R.06.APOE-4isnotassociatedwithAlzheimer'sdiseaseinelderlyNigerians.AnnalsofNeurology,59,182-185.Gustafson,L.,Brun,A.,Englund,E.,Hagnell,O.,Nilsson,K.,Stensmyr,M.,Ohlin,A.K.,andAbrahamson,M.98.A50-yearperspectiveofafamilywithchromosome-14-linkedAlzheimer'sdisease.HumanGenetics,102,253-257.Guttman,L.54.Somenecessaryconditionsforcommonfactoranalysis.Psy-chometrika,19,149-161.Haas,C.,andDeStrooper,B.999.ThepresenilinsinAlzheimer'sdisease:Pro-teoloysisholdsthekey.Science,286,916-919.Habeck,C.,Foster,N.L.,Perneczky,R.,Kurz,A.,Alexopoulos,P.,Koeppe,R.A.,Drzezga,A.,andStern,Y.008.MultivariateandunivariateneuroimagingbiomarkersofAlzheimer'sdisease.NeuroImage,40,1503-1515.Hall,C.S.934.Emotionalbehaviorintherat.I.Defecationandurinationasmeasuresofindividualdierencesinemotionality.JournalofComparativeandPhysiologicalPsychology,18,385-403.Hall,C.S.36.Emotionalbehaviorintherat.III.Therelationshipbetweenemotionalityandambulatoryactivity.JournalofComparativePsychology,22,345-452.Hampton,R.R.,andSchwartz,B.L.004.Episodicmemoryinnonhumans:What,andwhere,iswhen?CurrentOpinioninNeurobiology,14,192-197.Han,M.-E.,Park,K.-H.,Baek,S.-Y.,Kim,B.-S.,Kim,J.-B.,Kim,H.-J.,andOh,S.-O.007.Inhibitoryeectsofcaeineonhippocampalneurogenesisandfunction.BiochemicalandBiophysicalResearchCommunications,356,976-980.Hanley,J.A.,andMcNeil,B.J.982.ThemeaninganduseoftheareaunderareceiveroperatingcharacteristicROCcurve.Radiology,143,29-36.269

PAGE 280

Hardy,J.997.TheAlzheimerfamilyofdiseases:Manyetiologies,onepathogen-esis?ProceedingsoftheNationalAcademyofSciencesUSA,94,2095-2097.Hardy,J.06.HastheamyloidcascadehypothesisforAlzheimer'sdiseasebeenproved?CurrentAlzheimersResearch,3,71-73.Hart,D.J.,Craig,D.,Compton,S.A.,Critchlow,S.,Kerrigan,B.M.,McIlroy,S.P.,andPassmore,A.P.03.AretrospectivestudyofthebehavioralandpsychologicalsymptomsofmidandlatephaseAlzheimer'sdisease.Interna-tionalJournalofGeriatricPsychiatry,18,1037-1042.Hartman,R.E.,Shah,A.,Fagan,A.M.,Schwetye,K.E.,Parsadanian,M.,Schul-man,R.N.,Finn,M.B.,andHoltzman,D.M.06.PomegranatejuicedecreasesamyloidloadandimprovesbehaviorinamousemodelofAlzheimer'sdisease.NeurobiologyofDisease,24,506-515.Hashimoto,M.,Rockenstein,E.,Crews,L.,andMasliah,E.03.RoleofproteinaggregationinmitochondrialdysfunctionandneurodegenerationinAlzheimer'sandParkinson'sdiseases.NeuromolecularMedicine,4,21-36.Hasselmo,M.E.,andWyble,B.P.997.Freerecallandrecognitioninanetworkmodelofthehippocampus.BehavioralBrainResearch,89,1-34.Hayashi,H.,Kimura,N.,Yamaguchi,H.,Hasegawa,K.,Yokoseki,T.,Shibata,M.,Yamamoto,N.,Michikawa,M.,Yoshikawa,Y.,Terao,K.,Matsuzaki,K.,Lemere,C.A.,Selkoe,D.J.,Naiki,H.,andYanagisawa,K.004.AseedforAlzheimeramyloidinthebrain.JournalofNeuroscience,24,4894-4902.Haykin,S.1999.NeuralNetworks:AComprehensiveFoundation,2ndEd.UpperSaddleRiver,NJ:PrenticeHall.Head,E.,Moat,K.,Das,P.,Sarsoza,F.,Poon,W.W.,Landsberg,G.,Cotman,C.W.,andMurphy,M.P.05.Beta-amyloiddepositionandtauphos-phorylationinclinicallycharacterizedagedcats.NeurobiologyofAging,26,749-763.Hebb,D.O.949.TheOrganizationofBehavior:ANeuropsychologicalTheory.NewYork:Wiley.Hebb,D.O.,andWilliams,K.A.46.Amethodforratinganimalintelligence.JournalofGeneralPsychology,34,59-65.Hebert,L.E.,Scherr,P.A.,Bienias,J.L.,Bennett,D.A.,andEvans,D.A.003.Alzheimer'sdiseaseintheU.S.Population:Prevalenceestimatesusingthe2000census.ArchivesofNeurology,60,1119-1122.270

PAGE 281

Heeren,D.J.,andCools,A.R.000.ClassifyingposturesoffreelymovingrodentswiththehelpofFourierdescriptorsandaneuralnetwork.BehavioralResearchMethods,Instrumentation,andComputers,32,56-62.Hepler,D.J.,Wenk,G.L.,Cribbs,B.L.,Olton,D.S.,andCoyle,J.T.985.Memoryimpairmentsfollowingbasalforebrainlesions.BrainResearch,346,8-14.Holcomb,L.A.,Dhanasekaran,M.,Hitt,A.R.,Young,K.A.,Riggs,M.,andManyam,B.V.006.BacopamonnieraextractreducesamyloidlevelsinPSAPPmice.JournalofAlzheimer'sDisease,9,243-251.Holcomb,L.A.,Gordon,M.N.,Jantzen,P.,Hsiao,K.,Du,K.,andMorgan,D.99.Behavioralchangesintransgenicmiceexpressingbothamyloidprecursorproteinandpresenilin-1mutations:Lackofassociationwithamyloiddeposits.BehaviorGenetics,29,177-185.Holcomb,L.,Gordon,M.,McGowan,E.,Yu,X.,Benkovic,S.,Jantzen,P.,Wright,K.,Saad,I.,Mueller,R.,Morgan,D.,Sanders,S.,Zehr,C.,OCampo,K.,Hardy,J.,Prada,C.,Eckman,C.,Younkin,S.,Hsaio,K.,andDu,K.1998.AcceleratedAlzheimer-typephenotypeintransgenicmicecarryingbothmu-tantamyloidprecursorproteinandpresenilin1transgenes.NatureMedicine,4,97-100.Holmes,C.,Wilkinson,D.,Dean,C.,Vethanayagam,S.,Olivieri,S.,Langley,A.,Pandita-Gunawardena,H.D.,Hogg,F.,Clare,C.,andDamms,J.2004.TheecacyofdonezepilinthetreatmentofneuropsychiatricsymptomsinAlzheimer'sdisease.Neurology,63,214-219.Holmstrom,L.,Koistinen,P.,Laaksonen,J.,andOja,E.97.Neuralandstatisticalclassiers:Taxonomyandtwocasestudies.IEEETransactionsonNeuralNetworks,8,5-17.Hopeld,J.J.982.Neuralnetworksandphysicalsystemswithemergentcollec-tivecomputationalabilities.ProceedingsoftheNationalAcademyofSciencesUSA,79,2554-2558.Hsiao,K.,Chapman,P.,Nilsen,S.,Eckman,C.,Harigaya,Y..,Younkin,S.,Yang,F.,andCole,G.996.Correlativememorydecits,Abetaelevation,andamyloidplaquesintransgenicmice.Science,274,99-102.Hughes,J.R.,Shanmugham,S.,Wetzel,L.C.,Bellur,S.,andHughes,C.A.989.TherelationshipbetweenEEGchangesandcognitivefunctionsindementia:AstudyinaVApopulation.ClinicalElectroencephalography,20,77-85.271

PAGE 282

Hyde,L.A.,Hoplight,B.J.,andDenenberg,V.H.998.Waterversionoftheradial-armmaze:Learninginthreeinbredstrainsofmice.BrainResearch,785,236-244.Ikeuchi,T.,andSisodia,S.S.02.Cell-freegenerationofthenotch1intracellulardomainNICDandAPP-CTfgamma:evidencefordistinctintramembranousgamma-secretase"activities.NeuromolecularMedicine,1,43-54.Ikonomovic,M.D.,Klunk,W.E.,Abrahamson,E.E.,Mathis,C.A.,Price,J.C.,Tsopelas,N.D.,Lopresti,B.J.,Ziolko,S.,Bi,W.,Paljug,W.R.,Debnath,M.L.,Hope,C.E.,Isanski,B.A.,Hamilton,R.L.,andDeKosky,S.T.2008.Post-mortemcorrelatesofinvivoPiB-PETamyloidimaginginatypicalcaseofAlzheimer'sdisease.Brain,131,1630-1645.Irizarry,M.,McNamara,M.,Fedorchak,K.,Hsiao,K.,andHyman,B.997.APPswtransgenicmicedevelopage-relatedAdepositsandneuropilabnor-malities,butnoneuronallossinCA1.JournalofNeuropathologyandExperi-mentalNeurology,56,965-973.Irwin,S.968.Comprehensiveobservationalassessment:Ia.Asystematic,quan-titativeprocedureforassessingthebehavioralandphysiologicalstateofthemouse.Psychopharmacologia,13,222-257.Izquierdo,J.A.,Costas,S.M.,Justel,E.A.,andRabiller,G.979.Eectofcaeineonthememoryofthemouse.Psychopharmacology,61,29-30.Jack,C.R.,Marjanska,M.,Wengenack,T.M.,Reyes,D.A.,Curran,G.L.,Lin,J.,Preboske,G.M.,Poduslo,J.F.,andGarwood,M.2007.MagneticresonanceimagingofAlzheimer'spathologyinthebrainsoflivingtransgenicmice:AnewtoolinAlzheimer'sdiseaseresearch.TheNeuroscientist,13,38-48.Jack,C.R.,Petersen,R.C.,Xu,Y.C.,Waring,S.C.,OBrien,P.C.,Tangalos,E.G.,Smith,G.E.,Ivnik,R.J.,andKokmen,E.97.MedialtemporalatrophyonMRIinnormalagingandverymildAlzheimer'sdisease.Neurology,49,786-794.Jacobs,D.,Sano,M.,Marder,K.,Bell,K.,Bylsma,F.,Laeche,G.,Albert,M.,Brandt,J.,andStern,Y.94.AgeatonsetofAlzheimer'sdisease:Relationtopatternofcognitivedysfunctionandrateofdecline.Neurology,44,1215-1220.Jaenisch,R.88.Transgenicanimals.Science,240,1468-1472.272

PAGE 283

Jankowsky,J.L.,Fadale,D.J.,Anderson,J.,Xu,G.M.,Gonzales,V.,Jenkins,N.A.,Copeland,N.G.,Lee,M.K.,Younkin,L.H.,Wagner,S.L.,Younkin,S.G.,andBorchelt,D.R.004.Mutantpresenilinsspecicallyelevatethelevelsofthe42residuebeta-amyloidpeptideinvivo:Evidenceforaugmentationofa42-specicgammasecretase.HumanMolecularGenetics,13,159-170.Jankowsky,J.L.,Younkin,L.H.,Gonzales,V.,Fadale,D.J.,Slunt,H.H.,Lester,H.A.,Younkin,S.G.,andBorchelt,D.R.007.RodentAmodulatesthesolubilityanddistributionofamyloiddepositsintransgenicmice.JournalofBiologicalChemistry,282,22707-22720.Janus,C.,andWestaway,D.2001.TransgenicmousemodelsofAlzheimer'sdisease.PhysiologyandBehavior,73,873-886.Jarrett,J.T.,andLansbury,P.T.93.Seedingonedimensionalcrystallization"ofamyloid:ApathogenicmechanisminAlzheimer'sdiseaseandscrapie?Cell,73,1055-1058.Jausovec,N.,andJausovec,K.2000.Dierencesinevent-relatedandinducedbrainoscillationsinthethetaandalphafrequencybandsrelatedtohumanintelligence.NeuroscienceLetters,293,191-194.Jee,S.W.,Cho,J.S.,Kim,C.K.,Hwang,D.Y.,Shim,S.B.,Lee,S.H.,Sin,J.S.,Kim,Y.S.,Park,J.H.,Lee,S.H.,Choi,S.Y.,andKim,Y.K.007.Anal-ysisofdierentiallyexpressedgenesinearly-andlate-stageAPPsw-transgenicandnormalmiceusingcDNAmicroarray.InternationalJournalofMolecularMedicine,19,461-468.Jee,S.W.,Cho,J.S.,Oh,J.H.,Shim,S.B.,Hwang,D.Y.,Lee,S.H.,Song,Y.S.,Lee,S.H.,andKim,Y.05.cDNAmicroarray-basedanalysisofdierentiallyexpressedgenesintransgenicbrainsexpressingNSE-controlledAPPsw.InternationalJournalofMolecularMedicine,16,547-552.Jelles,B.,vanBirgelen,J.H.,Slaets,J.P.,Hekster,R.E.,Jonkman,E.J.,andStam,C.J.999.Decreaseofnon-linearstructureintheEEGofAlzheimerpatientscomparedtohealthycontrols.ClinicalNeurophysiology,110,1159-1167.Jensen,M.T.,Mottin,M.D.,Cracchiolo,J.R.,Leighty,R.E.,andArendash,G.W.2005.Lifelongimmunizationwithhuman-amyloid-42protectsAlzheimer'stransgenicmiceagainstcognitiveimpairmentthroughoutaging.Neuroscience,130,667-684.Jeong,J.04.EEGdynamicsinpatientswithAlzheimer'sdisease.ClinicalNeurophysiology,115,1490-1505.273

PAGE 284

Jeong,J.,Gore,J.C.,andPeterson,B.S.01a.MutualinformationanalysisoftheEEGinpatientswithAlzheimer'sdisease.ClinicalNeurophysiology,112,827-835.Jeong,J.,Kim,S.Y.,andHan,S.-H.998.Non-lineardynamicalanalysisoftheEEGinAlzheimer'sdiseasewithoptimalembeddingdimension.Electroen-cephalographyandClinicalNeurophysiology,106,220-228.Jeong,J.,Kim,D.J.,Kim,S.Y.,Chae,J.H.,Go,H.J.,andKim,K.S.001b.EectoftotalsleepdeprivationonthedimensionalcomplexityofthewakingEEG.Sleep,24,197-202.Jitsumori,M.,Siemann,M.,Lehr,M.,andDelius,J.D.02.Anewapproachtotheformationofequivalenceclassesinpigeons.JournaloftheExperimentalAnalysisofBehavior,78,397-408.Johnson-Kozlow,M.,Kritz-Silverstein,D.,Barrett-Connor,E.,andMorton,D.002.Coeeconsumptionandcognitivefunctionamongolderadults.Amer-icanJournalofEpidemiology,156,842-850.Johnson-Wood,K.,Lee,M.,Motter,R.,Hu,K.,Gordon,G.,Barbour,R.,Khan,K.,Gordon,M.,Tan,H.,Games,D.,Lieberburg,I.,Schenk,D.,Seubert,P.,andMcConlogue,L.997.AmyloidprecursorproteinprocessingandA42depositioninatransgenicmousemodelofAlzheimer'sdisease.ProceedingsoftheNationalAcademyofSciencesUSA,94,1550-1555.Jones,R.N.,andGallo,J.J.000.DimensionsoftheMini-MentalStateExam-inationamongcommunitydwellingolderadults.PsychologicalMedicine,30,605-618.Jones,S.N.,andAyers,C.R.2006.PsychometricpropertiesandfactorstructureofanexpandedCERADneuropsychologicalbatteryinanelderlyVAsample.ArchivesofClinicalNeuropsychology,21,359-365.Jorgensen,P.,Bus,C.,Pallisgaard,N.,Bryder,M.,andJorgensen,A.L.996.FamilialAlzheimer'sdiseaseco-segregateswithaMet-146-Ilesubstitutioninpresenilin-1.ClinicalGenetics,50,281-286.Joseph,J.A.,Shukitt-Hale,B.,andCasadesus,G.005.Reversingthedeleteriouseectsofagingonneuronalcommunicationandbehavior:Benecialpropertiesoffruitpolyphenoliccompounds.AmericanJournalofClinicalNutrition,81,S313-S316.274

PAGE 285

Joseph,J.A.,Shukitt-Hale,B.,Denisova,N.A.,Prior,R.L.,Cao,G.,Martin,A.,Taglialatela,G.,andBickford,P.C.98.Long-termdietarystrawberry,spinach,orvitaminEsupplementationretardstheonsetofage-relatedneuronalsignal-transductionandcognitivebehavioraldecits.JournalofNeuroscience,18,8047-8055.Jost,M.C.,andGrossberg,G.T.996.TheevolutionofpsychiatricsymptomsinAlzheimer'sdisease:Anaturalhistorystudy.JournaloftheAmericanGeriatricSociety,44,1078-1081.Kafka,N.2003.ExtendingSEEforlarge-scalephenotypingofmouseopen-eldbehavior.BehaviorResearchMethods,Instruments,andComputers,35,294-301.Kaiser,H.F.58.Thevarimaxcriterionforanalyticrotationinfactoranalysis.Psychometrika,23,187-200.Kaiser,H.F.960.Theapplicationofelectroniccomputerstofactoranalysis.EducationalandPsychologicalMeasurement,20,141-151.Kalampokis,A.,Kotsavasiloglou,C.,Argyrakis,P.,andBaloyannis,S.003.Robustnessinbiologicalneuralnetworks.PhysicaA,317,581-590.Kandel,E.R.979.Smallsystemsofneurons.ScienticAmerican,2413,66-76.Kandel,E.R.001.Themolecularbiologyofmemorystorage:Adialoguebe-tweengenesandsynapses.Science,294,1030-1038.Kandel,E.R.,andPittenger,C.999.Thepast,thefutureandthebiologyofmemorystorage.PhilosophicalTransactionsoftheRoyalSocietyofLondon,SeriesB,354,2027-2052.Kang,J.,Lemaire,H.G.,Unterbeck,A.,Salbaum,J.M.,Masters,C.L.,Grzeschik,K.H.,Multhaup,G.,Beyreuther,K.,andMuller-Hill,B.1987.TheprecursorofAlzheimer'sdiseaseamyloidA4proteinresemblesacell-surfacereceptor.Nature,325,733-736.Kanne,S.M.,Balota,D.A.,Storandt,M.,McKeel,D.W.,Jr.,andMorris,J.C.998.RelatinganatomytofunctioninAlzheimer'sdisease:Neuropsycho-logicalprolespredictregionalneuropathology5yearslater.Neurology,50,979-985.Kanowski,S.,Hermann,W.M.,Stephan,K.,Wierich,W.,andHorr,R.996.ProofofecacyofGinkgobilobaspecialextractEGb761inoutpatientssuf-feringfrommildtomoderateprimarydegenerativedementiaoftheAlzheimertypeormulti-infarctdementia.Pharmacopsychiatry,29,47-56.275

PAGE 286

Kaplan,G.B.,Greenblatt,D.J.,Ehrenberg,B.L.,Goddard,J.E.,Cotreau,M.M.,Harmatz,B.A.,andShader,R.I.97.Dose-dependentpharmacokineticsandpsychomotoreectsofcaeineinhumans.JournalofClinicalPharmacol-ogy,37,693-703.Katzman,R.,Brown,T.,Fuld,P.,Peck,A.,Schechter,R.,andSchimmel,H.983.Validationofashortorientation-memory-concentrationtestofcogni-tiveimpairment.AmericanJournalofPsychiatry,140,734-739.Kaufer,D.,Cummings,J.andChristine,D.1996.EectoftacrineonbehavioralsymptomsinAlzheimer'sdisease:Anopen-labelstudy.JournalofGeriatricPsychiatryandNeurology,9,1-6.Kayed,R.,Head,E.,Thompson,J.L.,McIntire,T.M.,Milton,S.C.,Cotman,C.W.,andGlabe,C.G.03.Commonstructureofsolubleamyloidoligomersimpliescommonmechanismofpathogenesis.Science,300,486-489.Kelly,P.,Bondol,L.,Hunziker,D.,Schlecht,H.,Carver,K.,Maquire,E.,Abramow-ski,D.,Wiederhold,K.,Sturchler-Pierrat,C.,Jucker,M.,Bergmann,R.,Staufenbiel,M.,andSommer,B.03.Progressiveage-relatedimpairmentofcognitivebehaviorinAPP23transgenicmice.NeurobiologyofAging,24,365-378.Kelso,S.R.,Ganong,A.H.,andBrown,T.H.86.Hebbiansynapsesinhip-pocampus.ProceedingsoftheNationalAcademyofSciencesUSA,83,5326-5330.Kesner,R.P.,Gilbert,P.E.,andBarua,L.A.002.Theroleofthehippocampusinmemoryforthetemporalorderofasequenceofodors.BehavioralNeuro-science,116,286-290.Kiang,Y.M.003.Acomparativeassessmentofclassicationmethods.DecisionSupportSystems,35,441-454.King,D.,andArendash,G.002a.BehavioralcharacterizationoftheTg2576transgenicmodelofAlzheimer'sdiseasethrough19months.Physiology&Behavior,75,627-642.King,D.,andArendash,G.02b.MaintainedsynaptophysinimmunoreactivityinTg2576transgenicmiceduringaging:Correlationswithcognitiveimpair-ment.BrainResearch,926,58-68.King,D.L.,Arendash,G.W.,Crawford,F.,Sterk,T.,Menendez,J.,andMullan,M.J.99.Progressiveandgender-dependentcognitiveimpairmentintheAPPswtransgenicmousemodelforAlzheimer'sdisease.BehavioralBrainResearch,103,145-162.276

PAGE 287

Kirsch,P.,Besthorn,C.,Klein,S.,Rindeisch,J.,andOlbrich,R.2000.Thedi-mensionalcomplexityoftheEEGduringcognitivetasksreectstheimpairedinformationprocessinginschizophrenicpatients.InternationalJournalofPsy-chophysiology,36,237-246.Kitzawa,M.,Yamasaki,T.R.,andLaFerla,F.M.004.Microgliaasapotentialbridgebetweentheamyloidbeta-peptideandtau.AnnalsoftheNewYorkAcademyofSciences,1035,85-103.Klafki,H.W.,Staufenbiel,M.,Kornhuber,J.,andWiltfang,J.06.TherapeuticapproachestoAlzheimer'sdisease.Brain129,2840-2855.Klunk,W.E.,Engler,H.,Nordberg,A.,Bacskai,B.J.,Wang,Y.,Price,J.C.,Bergstrom,M.,Hyman,B.T.,Langstrom,B.,andMathis,C.A.2003.ImagingthepathologyofAlzheimer'sdisease:Amyloid-imagingwithpositronemissiontomography.NeuroimagingClinicsofNorthAmerica,13,781-789.Klunk,W.E.,Engler,H.,Nordberg,A.,Wang,Y.,Blomqvist,G.,Holt,D.P.,Bergstrom,M.,Savitcheva,I.,Huang,G.F.,Estrada,S.,Ausen,B.,Debnath,M.L.,Barletta,J.,Price,J.C.,Sandell,J.,Lopresti,B.J.,Wall,A.,Koivisto,P.,Antoni,G.,Mathis,C.A.,andLangstrom,B.04.Imagingbrainamy-loidinAlzheimer'sdiseasewithPittsburghCompound-B.AnnalsofNeurology,55,306-319.Knapp,M.J.,Knopman,D.S.,Solomon,P.R.,Pendlebury,W.W.,Davis,C.S.,andGracon,S.I.994.A30-weekrandomizedcontrolledtrialofhigh-dosetacrineinpatientswithAlzheimer'sdisease.JournaloftheAmericanMedicalAssociation,271,985-991.Knopman,D.,andNissen,M.J.1991.ProcedurallearningisimpairedinHunt-ington'sdisease:Evidencefromtheserialreactiontimetask.Neuropsychologia,29,245-254.Knott,V.,Engeland,C.,Mohr,E.,Mahoney,C.,andIlivitsky,V.000.AcutenicotineadministrationinAlzheimer'sdisease:AnexploratoryEEGstudy.Neuropsychobiology,41,210-220.Koenig,T.,Prichep,L.,Dierks,T.,Hubl,D.,Wahlund,L.O.,John,E.R.,andJelic,V.2005.DecreasedEEGsynchronizationinAlzheimer'sdiseaseandmildcognitiveimpairment.NeurobiologyofAging,26,165-171.Kogan,E.A.,Korczyn,A.D.,Virchovsky,R.G.,Klimovizky,S.S.,Treves,T.A.,andNeufeld,M.Y.001.EEGchangesduringlong-termtreatmentwithdonepezilinAlzheimer'sdiseasepatients.JournalofNeuralTransmission,108,1167-1173.277

PAGE 288

Koh,H.C.,andTan,G.005.Dataminingapplicationsinhealthcare.JournalofHealthcareInformationManagement,19,64-72.Kohonen,T.982.Self-organizedformationoftopologicallycorrectfeaturemaps.BiologicalCybernetics,43,59-69.Koizumi,K.,Nakajima,M.,Yuasa,S.,Saga,Y.,Sakai,T.,Kuriyama,T.,Shi-rasawa,T.,andKoseki,H.001.Theroleofpresenilin1duringsomitesegmentation.Development,128,1391-1402.Kolata,S.,Light,K.,Townsend,D.A.,Hale,G.,Grossman,H.C.,andMatzel,L.D.2005.Variationsinworkingmemorycapacitypredictindividualdier-encesingenerallearningabilitiesamonggeneticallydiversemice.NeurobiologyofLearningandMemory,84,241-246.Kolmogorov,A.N.957.Ontherepresentationofcontinuousfunctionsofseveralvariablesbysuperpositionofcontinuousfunctionsofonevariableandaddition.DokladyAkademeiiNaukSSSR,114,953-956.Kovacs,D.M.,Fausett,H.J.,Page,K.J.,Kim,T.-W.,Moir,R.D.,Merriam,D.E.,Hollister,R.D.,Hallmark,O.G.,Mancini,R.,Felsenstein,K.M.,Hyman,B.T.,Tanzi,R.E.,Wasco,W.996.Alzheimer-associatedpresenilins1and2:neuronalexpressioninbrainandlocalizationtointracellularmembranesinmammaliancells.NatureMedicine,2,224-229.Kowalik-Jankowska,T.,Ruta-Dolejsz,M.,Wisniewska,K.,Lanklewicz,L.,andKozlowski,H.2002.PossibleinvolvementofcopperIIinAlzheimerdisease.EnvironmentalHealthPerspectives,110Suppl.5,869-870.Kowalski,J.W.,Gawel,M.,Pfeer,A.,andBarcikowska,M.2001.ThediagnosticvalueofEEGinAlzheimer'sdisease:Correlationwiththeseverityofmentalimpairment.JournalofClinicalNeurophysiology,18,570-575.Kramer,A.F.,andErickson,K.I.07.Capitalizingoncorticalplasticity:Inu-enceofphysicalactivityoncognitionandbrainfunction.TrendsinCognitiveSciences,11,342-348.Kristensen,M.,andHansen,T.2004.Statisticalanalysesofrepeatedmeasuresinphysiologicalresearch:Atutorial.AdvancesinPhysiologyEducation,28,2-14.Kuncel,N.R.,Hezlett,S.A.,andOnes,D.S.004.Academicperformance,careerpotential,creativity,andjobperformance:Canoneconstructpredictthemall?JournalofPersonalityandSocialPsychology,86,148-161.LaFerla,F.M.006.AnarrayofgenesimplicatedinAlzheimer'sdisease.Neu-robiologyofAging,27,1078-1080.278

PAGE 289

LaFerla,F.M.,andOddo,S.005.Alzheimer'sdisease:Abeta,tauandsynapticdysfunction.TrendsinMolecularMedicine,11,170-176.Lalonde,R.,Dumont,M.,Staufenbiel,M.,Sturchler-Pierrat,C.,andStrazielle,C.2002.Spatiallearning,exploration,anxiety,andmotorcoordinationinfemaleAPP23transgenicmicewiththeSwedishmutation.BrainResearch,956,36-44.Lamour,Y.,Bassant,M.H.,Jobert,A.,andJoly,M.1989.Septo-hippocampalneuronsintheagedrat:Relationbetweentheirelectrophysiologicalandphar-macologicalpropertiesandbehavioralperformances.NeurobiologyofAging,10,181-186.Lance,C.E.,Butts,M.M.,andMichels,L.C.006.Thesourcesoffourcommonlyreportedcutocriteria:Whatdidtheyreallysay?OrganizationalResearchMethods,9,202-220.Lance,R.F.,Kennedy,M.L.,andLeberg,P.L.00.Classicationbiasindis-criminantfunctionanalysesusedtoevaluateputativelydierenttaxa.JournalofMammalogy,81,245-249.Landauer,T.K.1986.Howmuchdopeopleremember?Someestimatesofthequantityoflearnedinformationinlong-termmemory.CognitiveScience,10,477-493.Larder,B.,Wang,D.,Revell,A.,Montaner,J.,Harrigan,R.,DeWolf,F.,Lange,J.,Wegner,S.,Ruiz,L.,Perez-Elias,M.J.,Emery,S.,Gatell,J.,Monforte,A.D.,Torti,C.,Zazzi,M.,andLane,C.2007.ThedevelopmentofarticialneuralnetworkstopredictvirologicalresponsetocombinationHIVtherapy.AntiviralTherapy,12,15-24.Larson,E.,Wang,L.,Bowen,J.,McCormick,W.,Teri,L.,Crane,P.,andKukull,W.006.Exerciseisassociatedwithreducedriskforincidentdementiaamongpersons65yearsofageandolder.AnnalsofInternalMedicine,144,73-81.Larson,J.,Lynch,G.,Games,D.,andSeubert,P.999.Alterationsinsynaptictransmissionandlong-termpotentiationinhippocampalslicesfromyoungandagedPDAPPmice.BrainResearch,840,23-35.Lasko,T.A.,Bhagwat,J.G.,Zou,K.H.,andOhno-Machado,L.005.Theuseofreceiveroperatingcharacteristiccurvesinbiomedicalinformatics.JournalofBiomedicalInformatics,38,404-415.279

PAGE 290

Lazarov,O.,Robinson,J.,Tang,Y.-P.,Hairston,I.S.,Korade-Mirnics,Z.,Lee,V.M.-Y.,Hersh,L.B.,Sapolsky,R.M.,Mirnics,K.,andSisodia,S.S.2005.EnvironmentalenrichmentreducesAlevelsandamyloiddepositionintrans-genicmice.Cell,120,701-713.LeBars,P.L.,Katz,M.M.,Berman,N.,Itil,T.M.,Freedman,A.M.,andSchatzberg,A.F.1997.Aplacebo-controlleddouble-blind,randomizedtrialofanextractofGinkgobilobafordementia.NorthAmericanEGbstudygroup.JournaloftheAmericanMedicalAssociation,278,1327-1332.LeBlanc,A.C.05.TheroleofapoptoticpathwaysinAlzheimer'sdiseaseneurodegenerationandcelldeath.CurrentAlzheimer'sResearch,2,389-402.Ledbetter,D.H.,Riccardi,V.M.,Airhart,S.D.,Strobel,R.J.,Keenan,B.S.,andCrawford,J.D.81.Deletionsofchromosome15asacauseofthePrader-Willisyndrome.NewEnglandJournalofMedicine,304,325-329.Lee,C.-K.,Weindruch,R.,andProlla,T.A.00.Gene-expressionproleoftheagingbraininmice.NatureGenetics,25,294-297.Lee,M.K.,Borchelt,D.R.,Wong,P.C.,Sisodia,S.S.,andPrice,D.L.996.Transgenicmodelsofneurodegenerativediseases.CurrentOpinioninNeuro-biology,6,651-660.Leighty,R.,Nilsson,L.,Potter,H.,Costa,D.,Low,M.,Bales,K.Paul,S.,andArendash,G.004.UseofmultimetricstatisticalanalysistocharacterizeanddiscriminatebetweentheperformanceoffourAlzheimer'stransgenicmouselinesdieringinAbetadeposition.BehaviouralBrainResearch,153,107-121.Leighty,R.E.03.Correlation,discriminantfunction,andfactoranalysesofabehavioralassessmentbatteryfortransgenicityandstrain-dierencesinmice.Master'sthesis,UniversityofSouthFlorida.Leighty,R.E.,Runfeldt,M.J.,Berndt,D.J.,Schleif,W.S.,Cracchiolo,J.R.,Potter,H.,andArendash,G.W.2008.UseofarticialneuralnetworkstodeterminecognitiveimpairmentandtherapeuticeectivenessinAlzheimer'stransgenicmice.JournalofNeuroscienceMethods,167,358-366..Lesne,S.,Koh,M.T.,Kotilinek,L.,Kayed,R.,Glabe,C.G.,Yang,A.,Gallagher,M.,andAshe,K.H.006.Aspecicamyloid-betaproteinassemblyinthebrainimpairsmemory.Nature,440,352-357.Letemendia,F.,andPampiglione,G.58.ClinicalandelectroencephalographicobservationsinAlzheimer'sdisease.JournalofNeurology,NeurosurgeryandPsychiatry,21,167-172.280

PAGE 291

Letenneur,L.,Larrieu,S.,andBarberger-Gateau,P.04.Alcoholandtobaccoconsumptionasriskfactorsofdementia:Areviewofepidemiologicalstudies.BiomedicineandPharmacotherapy,58,95-99.Leung,H.-C.,Skudlarski,P.,Gatenby,J.C.,Peterson,B.S.,andGore,J.C.000.Anevent-relatedfunctionalMRIstudyoftheStroopcolorwordinterferencetask.CerebralCortex,10,552-560.Levine,M.S.,Lloyd,R.L.,Fisher,R.S.,Hull,C.D.,andBuchwald,N.A.987.Sensory,motorandcognitivealterationsinagedcats.NeurobiologyofAging,8,253-263.Levy-Lahad,E.,Poorkaj,P.,Wang,K.,Fu,Y.H.,Oshima,J.,Mulligan,J.,andSchellenberg,G.D.96.GenomicstructureandexpressionofSTM2,thechromosome1familialAlzheimerdiseasegene.Genomics,34,198-204.Lewin,B.08.GenesIX.Sudbury,MA:JonesandBartlettPublishers,Inc.Lilliefors,H.W.68.OntheKolmogorov-Smirnovtestfornormalitywithmeanandvarianceunknown.JournaloftheAmericanStatisticalAssociation,62,399-402.Lim,G.P.,Chu,T.,Yang,F.,Beech,W.,Frautschy,S.A.,andCole,G.M.001.ThecurryspicecurcuminreducesoxidativedamageandamyloidpathologyinanAlzheimertransgenicmouse.JournalofNeuroscience,21,8370-8377.Lin,F.H.,Lin,R.,Wisniewski,H.M.,Hwang,Y.-W.,Grundke-Iqbal,I.,Healy-Louie,G.,andIqbal,K.92.Detectionofpointmutationsincodon331ofmitochondrialNADHdehydrogenasesubunit2inAlzheimer'sbrains.Bio-chemical&BiophysicalResearchCommunications,182,238-246.Lindsay,J.,Laurin,D.,Verreault,R.,Hebert,R.,Helliwell,B.,Hill,G.B.,andMcDowell,I.02.RiskfactorsforAlzheimer'sdisease:Aprospectiveanal-ysisfromtheCanadianStudyofHealthandAging.AmericanJournalofEpidemiology,156,445-453.Lipton,S.A.005.ThemolecularbasisofmemantineactioninAlzheimer'sdis-easeandotherneurologicaldisorders:Low-anity,uncompetitiveantagonism.CurrentAlzheimerResearch,2,155-165.Lisboa,P.J.,andTaktak,A.F.G.2006.Theuseofarticialneuralnetworksindecisionsupportincancer:Asystematicreview.NeuralNetworks,19,408-415.Lister,R.G.987.Theuseofaplus-mazetomeasureanxietyinthemouse.Psychopharmacology,92,180-185.281

PAGE 292

Liu,B.,Jiang,T.,Ma,S.,Zhao,H.,Li,J.,Jiang,X.,andZhang,J.006.Exploringcandidategenesforhumanbraindiseasesfromabrain-specicgenenetwork.BiochemicalandBiophysicalResearchCommunications,349,1308-1314.Lleo,A.,Blesa,R.,Gendre,J.,Castellvi,M.,Pastor,P.,Queralt,R.,andOliva,R.01.Anovelpresenilin2genemutationD439Ainapatientwithearly-onsetAlzheimer'sdisease.Neurology,57,1926-1928.Locascio,J.J.,Growdon,J.H.,andCorkin,S.995.Cognitivetestperformanceindetecting,staging,andtrackingAlzheimer'sdisease.ArchivesofNeurology,52,1087-1099.Loewenstein,D.A.,Acevedo,A.,Luis,C.,Crum,T.,Barker,W.W.,andDuara,R.004.SemanticinterferencedecitsandthedetectionofmildAlzheimer'sdiseaseandmildcognitiveimpairmentwithoutdementia.JournaloftheIn-ternationalNeuropsychologicalSociety,10,91-100.Loewenstein,D.A.,Acevedo,A.,Schram,L.,Ownby,R.,White,G.,Mogosky,B.,Barker,W.W.,andDuara,R.003.SemanticinterferenceinmildAlzheimer'sdisease:Preliminaryndings.AmericanJournalofGeriatricPsy-chiatry,11,252-255.Loewenstein,D.A.,Amigo,A.,Duara,R.,Gutterman,A.,Hurwitz,D.,Berkowitz,N.,Wilkie,F.,Weinberg,G.,Black,B.,Gittelman,B.,andEisdorfer,C.1989.AnewscalefortheassessmentoffunctionalstatusinAlzheimer'sdiseaseandrelateddisorders.JournalofGerontology,44,114-121.Loewenstein,D.A.,Arguelles,T.,Barker,W.W.,Schram,L.,Ownby,R.,Acevedo,A.,Mogosky,B.,White,G.,andDuara,R.001.TheutilityofamodiedobjectmemorytestindistinguishingbetweenthreedierentagegroupsofAlzheimer'sdiseasepatientsandnormalcontrols.JournalofMentalHealthandAging,7,317-324.Loewenstein,D.A.,Rubert,M.P.,Arguelles,T.,andDuara,R.1995.Neu-ropsychologicaltestperformanceandpredictionoffunctionalcapacitiesamongSpanish-speakingandEnglish-speakingpatientswithdementia.ArchivesofClinicalNeuropsychology,10,75-88.Logue,S.F.,Paylor,R.,andWehner,J.M.97.HippocampallesionscauselearningdecitsininbredmiceintheMorriswatermazeandconditioned-feartask.BehavioralNeuroscience,111,104-113.Lopez-Rangel,E.,andLewis,M.2006.Loudandclearevidenceforgenesilencingbyepigeneticmechanismsinautismspectrumandrelatedneurodevelopmentaldisorders.ClinicalGenetics,69,21-25.282

PAGE 293

Lutzenberger,W.,Birbaumer,N.,Flor,H.,Rockstroh,B.,andElbert,T.992.DimensionalanalysisofthehumanEEGandintelligence.NeuroscienceLetters,143,10-14.Lynch,M.A.004.Long-termpotentiationandmemory.PhysiologicalReviews,84,87-136.MacCallum,R.C.,Widaman,K.F.,Zhang,S.,andHong,S.999.Samplesizeinfactoranalysis.PsychologicalMethods,4,84-99.Maccioni,R.,Munoz,J.andBarbeito,L.01.ThemolecularbasesofAlzheimer'sdiseaseandotherneurodegenerativedisorders.ArchivesofMedicalResearch,32,367-381.Magenis,R.E.,Brown,M.G.,Lacy,D.A.,Budden,S.,andLaFranchi,S.87.IsAngelmansyndromeanalternateresultofdel15q11q13?AmericanJournalofMedicalGenetics,28,829-838.Maia,L.,andDeMendonca,A.2002.DoescaeineintakeprotectfromAlzheimer'sdisease?EuropeanJournalofNeurology,9,377-382.Markesbery,W.R.997.NeuropathologicalcriteriaforthediagnosisofAlzheimer'sdisease.NeurobiologyofAging,18S4,S13-S19.Markowska,A.L.,Stone,W.S.,Ingram,D.K.,Reynolds,J.,Gold,P.E.,Conti,L.H.,Pontecorvo,M.J.,Wenk,G.L.,andOlton,D.S.89.Individualdierencesinaging:Behavioralandneurobiologicalcorrelates.NeurobiologyofAging,10,31-43.Marnett,L.J.,andKalgutkar,A.S.99.Cyclooxygenase2inhibitors:discovery,selectivityandthefuture.TrendsPharmacologicalSciences,20,465469.Martone,M.,Butters,N.,Payne,M.,Becker,J.,andSax,D.84.Dissocia-tionsbetweenskilllearningandverbalrecognitioninamnesiaanddementia.ArchivesofNeurology,41,965-970.Masaki,K.H.,Losonczy,K.G.,andIzmirlian,G.00.AssociationofvitaminEandCsupplementusewithcognitivefunctionanddementiainelderlymen.Neurology,54,1265-1272.Masliah,E.,Sisk,A.,Mallory,M.,andGames,D.01.Neurobrillarypathol-ogyintransgenicmiceoverexpressingV717Fbeta-amyloidprecursorprotein.JournalofNeuropathologyandExperimentalNeurology,60,357-368.283

PAGE 294

Masliah,E.,Sisk,A.,Mallory,M.,Mucke,L.,Schenk,D.,andGames,D.996.ComparisonofneurodegenerativepathologyintransgenicmiceoverexpressingV717Fbeta-amyloidprecursorproteinandAlzheimer'sdisease.JournalofNeuroscience,16,5795-5811.Masters,C.L.,Simms,G.,Weinman,N.A.,Multhaup,G.,McDonald,B.L.,andBeyreuther,K.85.AmyloidplaquecoreproteininAlzheimerdiseaseandDownsyndrome.ProceedingsoftheNationalAcademyofSciencesUSA,82,4245-4249.Mastropaolo,J.,Nadi,N.S.,Ostrowski,N.L.,andCrawley,J.N.88.Galaninantagonizesacetylcholineonamemorytaskinbasalforebrain-lesionedrats.ProceedingsoftheNationalAcademyofSciencesUSA,85,9841-9845.Mathis,C.A.,Wang,Y.,andKlunk,W.E.004.Imagingbeta-amyloidplaquesandneurobrillarytanglesintheaginghumanbrain.CurrentPharmaceuticalDesign,13,1469-1492.Mattson,M.P.004.PathwaystowardsandawayfromAlzheimer'sdisease.Nature,430,631-639.Mattson,M.P.,Chan,S.L.,andDuan,W.02.Modicationofbrainagingandneurodegenerativedisordersbygenes,diet,andbehavior.PhysiologicalReviews,82,637-672.Matzel,L.D.,Han,Y.R.,Grossman,H.,Karnik,M.S.,Patel,D.,Scott,N.,Specht,S.M.,andGandhi,C.C.03.Individualdierencesintheexpressionofageneral"learningabilityinmice.JournalofNeuroscience,23,6423-6433.Maurer,K.,Volk,S.,andGerbaldo,H.97.AugusteD.andAlzheimer'sdisease.Lancet,349,1546-1549.Maviel,T.,Durkin,T.P.,Menzaghi,F.,andBontempi,B.04.Sitesofneocor-ticalreorganizationcriticalforremotespatialmemory.Science,305,96-99.McClearn,G.E.,Johansson,B.,Berg,S.,Pedersen,N.L.,Ahern,F.,Petrill,S.A.,andPlomin,R.97.Substantialgeneticinuenceoncognitiveabilitiesintwins80ormoreyearsold.Science,276,1560-1563.McCulloch,W.S.,andPitts,W.43.Alogicalcalculusoftheideasimmanentinnervousactivity.BulletinofMathematicalBiophysics,5,115-133.McDonald,M.P.,andOvermier,J.B.98.Presentimperfect:Acriticalre-viewofanimalmodelsofthemnemonicimpairmentsinAlzheimer'sdisease.NeuroscienceandBiobehavioralReviews,22,99-120.284

PAGE 295

McGeer,P.L.,andMcGeer,E.G.07.NSAIDsandAlzheimer'sdisease:Epi-demiological,animalmodelandclinicalstudies.NeurobiologyofAging,28,639-647.McGeer,P.L.,Schulzer,M.,andMcGeer,E.G.96.Arthritisandanti-inammatoryagentsaspossibleprotectivefactorsforAlzheimer'sdisease:areviewof17epidemiologicstudies.Neurology,47,425432.McKhann,G.,Drachman,D.,Folstein,M.,Katzman,R.,Price,D.,andStadlan,E.M.984.ClinicaldiagnosisofAlzheimer'sdisease:ReportoftheNINCDS-ADRDAWorkGroupundertheauspicesofDepartmentofHealthandHumanServicesTaskForceonAlzheimer'sDisease.Neurology,34,939-944.Meyer,D.E.,andSchvaneveldt,R.W.1971.Facilitationinrecognizingpairsofwords:Evidenceofadependencebetweenretrievaloperations.JournalofExperimentalPsychology,90,227-234.Meyer-Luehmann,M.,Coomaraswamy,J.,Bolmont,T.,Kaeser,S.,Schaefer,C.,Kilger,E.,Neuenschwander,A.,Abramowski,D.,Frey,P.,Jaton,A.L.,Vigouret,J.-M.,Paganetti,P.,Walsh,D.M.,Mathews,P.M.,Ghiso,J.,Staufenbiel,M.,Walker,L.C.,andJucker,M.06.Exogenousinductionofcerebralbeta-amyloidogenesisisgovernedbyagentandhost.Science,313,1781-1784.Middei,S.,Daniele,S.,Caprioli,A.,Ghirardi,O.,andAmmassari-Teule,M.006.ProgressivecognitivedeclineinatransgenicmousemodelofAlzheimer'sdiseaseoverexpressingmutanthAPPswe.Genes,BrainandBehavior,5,249-256.Milgram,N.W.,Head,E.,Zicker,S.C.,Ikeda-Douglas,C.J.,Murphey,H.,Muggenburg,B.,Siwak,C.,Tapp,D.,andCotman,C.W.05.Learningabilityinagedbeagledogsispreservedbybehavioralenrichmentanddietaryfortication:Atwo-yearlongitudinalstudy.NeurobiologyofAging,26,77-90.Miller,A.S.,Blott,B.H.,andHames,T.K.92.Reviewofneuralnetworkapplicationsinmedicalimagingandsignalprocessing.MedicalandBiologicalEngineeringandComputing,30,449-464.Miller,G.A.1956.Themagicalnumberseven,plusorminustwo:Somelimitsonourcapacityforprocessinginformation.PsychologicalReview,63,81-97.Miller,R.M.,andFedero,H.J.008.Isoform-speciceectsofApoEonHSVimmediateearlygeneexpressionandestablishmentoflatency.NeurobiologyofAging,29,71-77.Minkeviciene,R.,Banerjee,P.,andTanila,H.2004.Memantineimprovesspa-tiallearninginatransgenicmousemodelofAlzheimer'sdisease.JournalofPharmacologyandExperimentalTherapeutics,311,677-682.285

PAGE 296

Minoshima,S.003.ImagingAlzheimer'sdisease:Clinicalapplications.Neu-roimagingClinicsofNorthAmerica,13,769-780.Minsky,M.,andPapert,S.69.Perceptrons:AnIntroductiontoComputationalGeometry.Cambridge,MA:MITPress.Mirakur,A.,Craig,D.,Hart,D.J.,McIlroy,S.P.,andPassmore,A.P.004.BehavioralandpsychologicalsyndromesinAlzheimer'sdisease.InternationalJournalofGeriatricPsychiatry,19,1035-1039.Mirnics,K.,andPevsner,J.004.Progressintheuseofmicroarraytechnologytostudytheneurobiologyofdisease.NatureNeuroscience,7,434-439.Mitchell,D.B.989.Howmanymemorysystems?Evidencefromaging.JournalofExperimentalPsychology:Learning,Memory,andCognition,15,31-49.Miyake,A.,Friedman,N.P.,Rettinger,D.A.,Shah,P.,andHegarty,M.001.Howarevisuospatialworkingmemory,executivefunctioning,andspatialabil-itiesrelated?Alatent-variableanalysis.JournalofExperimentalPsychology:General,130,621-640.Miyashita,Y.04.Cognitivememory:Cellularandnetworkmachineriesandtheirtop-downcontrol.Science,306,435-440.Moechars,D.,Lorent,K.,DeStrooper,B.,Dewachter,I.,andVanLeuven,F.996.Expressioninbrainofamyloidprecursorproteinmutatedinthealpha-secretasesitecausesdisturbedbehavior,neuronaldegenerationandprematuredeathintransgenicmice.EMBOJournal,15,1265-1274.Mohajeri,M.,Madani,R.,Saini,K.,Lipp,H.,Nitsch,R.,andWolfer,D.004.Theimpactofgeneticbackgroundonneurodegenerationandbehaviorinseizuremice.Genes,BrainandBehavior,3,228-239.Molle,M.,Marshall,L.,Wolf,B.,Fehm,H.L.,andBorn,J.99.EEGcom-plexityandperformancemeasuresofcreativethinking.Psychophysiology,36,95-104.Montgomery,K.C.53.Concerningtheuseofanalysisofvarianceonlatencydata.AmericanJournalofPsychology,66,131-135.Montgomery,K.C.955.Therelationbetweenfearinducedbynovelstimulationandexploratorybehavior.JournalofComparativeandPhysiologicalPsychol-ogy,48,254-260.Morelli,L.,Prat,M.I.,Levy,E.,Mangone,C.A.,Castano,E.M.1998,Prese-nilin1met146leuvariantduetoanA-Ttransversioninanearly-onsetfamilialAlzheimer'sdiseasepedigreefromArgentina.Genetics,53,469-473.286

PAGE 297

Morihara,T.,Chu,T.,Ubeda,O.,Beech,W.,andCole,G.M.002.Selec-tiveinhibitionofAbeta42productionbyNSAIDR-enantiomers.JournalofNeurochemistry,83,1009-1012.Morris,J.C.,Edland,S.,Clark,C.,Galasko,D.,Koss,E.,Mohs,R.,vanBelle,G.,Fillenbaum,G.,andHeyman,A.93.TheConsortiumtoEstablishaReg-istryforAlzheimer'sDiseaseCERAD:PartIV.RatingsofcognitivechangeinthelongitudinalassessmentofprobableAlzheimer'sdisease.Neurology,43,2457-2465.Morris,J.C.,Heyman,A.,Mohs,R.C.,Hughes,J.P.,vanBelle,G.,Fillenbaum,G.,Mellits,E.D.,andClark,C.989.TheconsortiumtoestablisharegistryforAlzheimer'sdiseaseCERAD.Part1.ClinicalandneuropsychologicalassessmentofAlzheimer'sdisease.Neurology,39,1159-1165.Morris,M.,Evans,D.,Tangney,C.,Bienias,J.,andWilson,R.005.Fishcon-sumptionandcognitivedeclinewithageinalargecommunitystudy.ArchivesofNeurology,62,1849-1853.Morris,M.C.,Beckett,L.A.,Scherr,P.A.,Hebert,L.E.,Bennett,D.A.,Field,T.S.,andEvans,D.A.998.VitaminEandvitaminCsupplementuseandriskofincidentAlzheimerdisease.AlzheimerDiseaseandAssociatedDisorders,12,121-126.Morris,R.,Anderson,E.,Lynch,G.,andBaudry,M.986.Selectiveimpairmentoflearningandblockadeoflong-termpotentiationbyanN-methyl-D-aspartatereceptorantagonist,AP5.Nature,319,774-776.Morris,R.G.M.1981.Spatiallocalizationdoesnotdependonthepresenceoflocalcues.LearningandMotivation,12,239-260.Morris,R.G.M.1984.Developmentsofwater-mazeprocedureforstudyingspa-tiallearningintherat.JournalofNeuroscienceMethods,11,47-60.Morris,R.G.M.001.Episodic-likememoryinanimals:Psychologicalcriteria,neuralmechanismsandthevalueofepisodic-liketaskstoinvestigateanimalmodelsofneurodegenerativedisease.PhilosophicalTransactionsoftheRoyalSocietyofLondon,SeriesB,356,1453-1465.Morris,R.G.M.,Garrud,P.,Rawlins,J.N.P.,andO'Keefe,J.82.Placenavigationimpairedinratswithhippocampallesions.Nature,297,681-683.Morrison,J.H.,andHof,P.R.997.Lifeanddeathofneuronsintheagingbrain.Science,278,412-419.287

PAGE 298

Moser,V.C.,Cheek,B.M.,andMacPhail,R.C.95.Amultidisciplinaryapproachtotoxicologicalscreening.III.Neurobehavioraltoxicology.JournalofToxicologyandEnvironmentalHealth,45,173-210.Moulin,C.J.A.,Laine,M.,Rinne,J.O.,Kaasinen,V.,Sipila,H.,Hiltunen,J.,andKangasmaki,A.07.Brainfunctionduringmulti-triallearninginmildcognitiveimpairment:APETactivationstudy.BrainResearch,1136,132-141.Mudher,A.andLovestone,S.02.Alzheimer'sdisease:Dotauistsandbaptistsnallyshakehands?TRENDSinNeurosciences,25,22-26.Mullan,M.,Crawford,F.,Axelman,K.,Houlden,H.,Lilius,L.,Winblad,B.,andLannfelt,L.992.ApathogenicmutationforprobableAlzheimer'sdiseaseintheAPPgeneattheN-terminusofbeta-amyloid.NatureGenetics,1,345-347.Mundt,J.C.,Freed,D.M.,andGreist,J.H.000.Layperson-basedscreeningforearlydetectionofAlzheimer'sdisease:Developmentandvalidationofaninstrument.JournalofGerontology:PsychologicalSciences,55B,P163-P170.Murrell,J.,Farlow,M.,Ghetti,B.,andBenson,M.991.Amutationintheamy-loidprecursorproteinassociatedwithhereditaryAlzheimer'sdisease.Science,254,97-99.Nair,J.,Nair,S.S.,Kashani,J.H.,Reid,J.C.,Mistry,S.I.,andVargas,V.G.999.Analysisofthesymptomsofdepression:Aneuralnetworkapproach.PsychiatryResearch,87,193-201.Nakayama,H.,Uchida,K.,andDoi,K.004.Acomparativestudyofage-relatedbrainpathology:Areneurodegenerativediseasespresentinnonhumananimals?MedicalHypotheses,63,198-202.Nalini,K.,Karanth,K.S.,Rao,A.,andAroor,A.R.95.EectsofCelastruspaniculatusonpassiveavoidanceperformanceandbiogenicamineturnoverinalbinorats.JournalofEthnopharmacology,47,101-108.Neely,R.77.Discriminantanalysisforpredictionofcollegegraduation.Edu-cationalandPsychologicalMeasurement,37,965-970.Nehlig,A.,Daval,J.-L.,andDebry,G.992.Caeineandthecentralnervoussystem:Mechanismsofaction,biochemical,metabolicandpsychostimulanteects.BrainResearchReviews,17,139-170.Nelson,P.T.,Greenberg,S.G.,andSaper,C.B.94.Neurobrillarytanglesinthecerebralcortexofsheep.NeuroscienceLetters,170,187-190.288

PAGE 299

Neyman,J.,andPearson,E.S.1933.Thetestingofstatisticalhypothesesinrelationtoprobabilitiesapriori.Reprintedin:Neyman,J.,andPearson,E.S.967.JointStatisticalPapers.Cambridge:CambridgeUniversityPress.pp.186-202.Nguyen,P.V.,Abel,T.,Kandel,E.R.,andBourtchouladze,R.000.Strain-dependentdierencesinLTPandhippocampus-dependentmemoryininbredmice.LearningandMemory,7,170-179.Nichol,K.E.,Parachikova,A.I.,andCotman,C.W.007.ThreeweeksofrunningwheelexposureimprovescognitiveperformanceintheagedTg2576mouse.BehavioralBrainResearch,184,124-132.Nilsson,J.,Ohlsson,M.,Thulin,L.,Hoglund,P.,Nashef,S.A.M.,andBrandt,J.006.Riskfactoridenticationandmortalitypredictionincardiacsurgeryusingarticialneuralnetworks.JournalofThoracicandCardiovascularSur-gery,132,12-19.Nilsson,L.N.,Arendash,G.W.,Leighty,R.E.,Costa,D.A.,Low,M.A.,Garcia,M.F.,Cracchiolo,J.R.,Rojiani,A.,Wu,X.,Bales,K.R.,Paul,S.M.,andPotter,H.2004.CognitiveimpairmentinPDAPPmicedependsonApoEandACT-catalyzedamyloidformation.NeurobiologyofAging,25,1153-1167.Nissen,M.J.,andBullemer,P.987.Attentionalrequirementsoflearning:Evi-dencefromperformancemeasures.CognitivePsychology,19,1-32.Noble,W.S.06.Whatisasupportvectormachine?NatureBiotechnology,24,1565-1567.Nunn,J.A.,Graydon,F.J.,Polkey,C.E.,andMorris,R.G.999.Dierentialspatialmemoryimpairmentafterrighttemporallobectomydemonstratedusingtemporaltitration.Brain,122,47-59.Nyberg,L.,Habib,R.,andHerlitz,A.000.Brainactivationduringepisodicmemoryretrieval:Sexdierences.ActaPsychologica,105,181-194.Obuchowski,N.A.03.Receiveroperatingcharacteristiccurvesandtheiruseinradiology.Radiology,229,3-8.Oddo,S.,Caccamo,A.,Kitazawa,M.,Tseng,B.P.,andLaFerla,F.M.03a.AmyloiddepositionprecedestangleformationinatripletransgenicmodelofAlzheimer'sdisease.NeurobiologyofAging,24,1063-1070.Oddo,S.,Caccamo,A.,Shepherd,J.D.,Murphy,M.P.,Golde,T.E.,Kayed,R.,Metherate,R.,Mattson,M.P.,Akbari,Y.,andLaFerla,F.M.03b.Triple-transgenicmodelofAlzheimer'sdiseasewithplaquesandtangles:IntracellularAbetaandsynapticdysfunction.Neuron,39,409-421.289

PAGE 300

Oddo,S.,Caccamo,A.,Tran,L.,Lambert,M.P.,Glabe,C.G.,Klein,W.L.,andLaFerla,F.M.06.Temporalproleofamyloid-betaAbetaoligomeriza-tioninaninvivomodelofAlzheimer'sdisease.AlinkbetweenAbetaandtaupathology.JournalofBiologicalChemistry,281,1599-1604.O'Donnell,B.F.,Drachman,D.A.,Lew,R.A.,andSwearer,J.M.88.Mea-suringdementia:Assessmentofmultipledecitdomains.JournalofClinicalPsychology,44,916-923.Olazaran,J.,Muniz,R.,Reisberg,B.,Pena-Casanova,J.,delSer,T.,Cruz-Jentoft,A.,Serrano,P.,Navarro,E.,GarciadelaRocha,M.,Frank,A.,Galiano,M.,Fernandez-Bullido,Y.,Serra,J.,Gonzalez-Salvador,M.,andSevilla,C.004.Benetsofcognitive-motorinterventioninMCIandmildtomoderateAlzheimerdisease.Neurology,63,2348-2353.Olson,J.M.,Goddard,K.A.B.,andDudek,D.M.002.Asecondlocusforvery-late-onsetAlzheimerdisease:agenomescanrevealslinkageto20pandepistasisbetween20pandtheamyloidprecursorproteinregion.AmericanJournalofHumanGenetics,71,154-161.Olton,D.S.977.Spatialmemory.ScienticAmerican,2366,82-98.Olton,D.S.,andSamuelson,R.J.76.Remembranceofplacespassed:Spa-tialmemoryinrats.JournalofExperimentalPsychology:AnimalBehaviorProcesses,2,97-116.Ono,K.,Hasegawa,K.,Naiki,H.,andYamada,M.004.Curcuminhaspotentanti-amyloidogeniceectsforAlzheimer'sbeta-amyloidbrilsinvitro.JournalofNeuroscienceResearch,75,742-750.Ossenkopp,K.P.,Sorenson,L.,andMazmanian,D.S.994.Factoranaly-sisofopeneldbehaviorintheratRattusnorvegicus:ApplicationofthePARAFACmodeltoalongitudinaldataset.BehavioralProcesses,31,129-144.Ownby,R.L.,Loewenstein,D.A.,Schram,L.,andAcevedo,A.004.AssessingthecognitiveabilitiesthatdierentiatepatientswithAlzheimer'sdiseasefromnormals:Singleandmultiplefactormodels.InternationalJournalofGeriatricPsychiatry,19,232-242.Papassotiropoulos,A.,Wollmer,M.A.,Aguzzi,A.,Hock,C.,Nitsch,R.M.,anddeQuervain,D.J.2005.Thepriongeneisassociatedwithhumanlong-termmemory.HumanMolecularGenetics,14,2241-2246.290

PAGE 301

Papassotiropoulos,A.,Stephan,D.A.,Huentelman,M.J.,Hoerndli,F.J.,Craig,D.W.,Pearson,J.V.,Huynh,K.-D.,Brunner,F.,Corneveaux,J.,Osborne,D.,Wollmer,M.A.,Aerni,A.,Coluccia,D.,Hanggi,J.,Mondadori,R.A.,Buchmann,A.,Reiman,E.M.,Caselli,R.J.,Henke,K.,anddeQuervain,D.J.-F.2006.CommonKibraallelesareassociatedwithhumanmemoryperformance.Science,314,475-478.Papik,K.,Molnar,B.,Schaefer,R.,Dombovari,Z.,Tulassay,Z.,andFeher,J.996.Applicationofneuralnetworksinmedicine:Areview.MedicalScienceMonitor,4,538-546.Pappas,B.A.,Bayley,P.J.,Bui,B.K.,Hansen,L.A.,andThal,L.J.000.CholineacetyltransferaseactivityandcognitivedomainscoresofAlzheimer'spatients.NeurobiologyofAging,21,11-17.Pardo,L.M.,andvanDuijn,C.M.005.Insearchofgenesinvolvedinneurode-generativedisorders.MutationResearch,592,89-101.Parsons,S.,andJones,G.00.Acousticidenticationoftwelvespeciesofecholo-catingbatbydiscriminantfunctionanalysisandarticialneuralnetworks.JournalofExperimentalBiology,203,2641-2656.Pascuale-Leone,A.,Wassermann,E.M.,Grafman,J.,andHallett,M.996.Theroleofthedorsolateralprefrontalcortexinimplicitprocedurallearning.ExperimentalBrainResearch,107,479-485.Patel,N.S.,Paris,D.,Mathura,V.,Quadros,A.N.,Crawford,F.C.,andMul-lan,M.J.05.InammatorycytokinelevelscorrelatewithamyloidloadintransgenicmousemodelsofAlzheimer'sdisease.JournalofNeuroinamma-tion,2,9.DOI:10.1186/1742-2094-2-9.Peek,M.S.,Russek-Cohen,E.,Wait,D.A.,andForseth,I.N.2002.Physiologicalresponsecurveanalysisusingnonlinearmixedmodels.Oecologia,132,175-180.Pendharkar,P.C.002.Acomputationalstudyontheperformanceofarti-cialneuralnetworksunderchangingstructuraldesignanddatadistribution.EuropeanJournalofOperationalResearch,138,155-177.Pennanen,C.,Kivipelto,M.,Tuomainen,S.,Hartikainen,P.,Hanninen,T.,Laakso,M.P.,Hallikainen,M.,Vanhanen,M.,Nissinen,A.,Helkala,E.L.,Vainio,P.,Vanninen,R.,Partanen,K.,andSoininen,H.2004.HippocampusandentorhinalcortexinmildcognitiveimpairmentandearlyAlzheimer'sdisease.NeurobiologyofAging,25,303-310.291

PAGE 302

Periani,D.,Bressi,S.,Cappa,S.F.,Vallar,G.,Alberoni,M.,Grassi,F.,Calt-agirone,C.,Ciplotti,L.,Franceschi,M.,Leniz,G.L.,andFazio,F.1993.Evidenceofmultiplememorysystemsinthehumanbrain.Brain,106,903-909.Perry,E.K.,Pickering,A.T.,Wang,W.W.,Houghton,P.J.,andPerry,N.S.1999.MedicinalplantsandAlzheimer'sdisease:Fromethnobotanytophytotherapy.JournalofPharmacyandPharmacology,51,527-534.Peterson,B.S.,Skudlarski,P.,Gatenby,J.C.,Zhang,H.,Anderson,A.W.,andGore,J.C.99.AnfMRIstudyofStroopword-colorinterference:Evidenceforcinguatesubregionssubservingmultipledistributedattentionalsystems.BiologicalPsychiatry,45,1237-1258.Peterson,L.R.,andPeterson,M.J.959.Short-termretentionofindividualitems.JournalofExperimentalPsychology,58,193-198.Pfeier,E.975.Ashortportablementalstatusquestionnairefortheassessmentoforganicbraindecitinelderlypatients.JournaloftheAmericanGeriatricSociety,23,433-441.Philipson,J.D.003.50yearsofmedicinalplantresearch{Everyprogressinmethodologyisaprogressisscience.PlantaMedica,69,491-495.Phillips-Wren,G.,Sharkey,P.,andDy,S.M.007.Mininglungcancerpatientdatatoassesshealthcareresourceutilization.ExpertSystemswithApplica-tions,35,1611-1619.Picciotto,M.R.,andWickman,K.998.Usingknockoutandtransgenicmicetostudyneurophysiologyandbehavior.PhysiologicalReview,78,1131-1163.Plomin,R.001.Thegeneticsofginhumanandmouse.NatureReviewsNeuro-science,2,136-141.Pompl,P.N.,Mullan,M.J.,Bjugstad,K.,andArendash,G.W.999.Adaptationofthecircularplatformspatialmemorytaskformice:UseindetectingcognitiveimpairmentintheAPPswtransgenicmousemodelforAlzheimer'sdisease.JournalofNeuroscienceMethods,87,87-95.Posner,M.I.,DiGirolamo,G.J.,andFernandez-Duque,D.997.Brainmecha-nismsofcognitiveskills.ConsciousnessandCognition,6,267-290.Postman,L.,andUnderwood,B.J.973.Criticalissuesininterferencetheory.MemoryandCognition,1,19-40.292

PAGE 303

Potter,H.,Wefes,I.andNilsson,L.001.Theinammation-inducedpatholog-icalchaperonesACTandapo-EarenecessarycatalystsofAlzheimeramyloidformation.NeurobiologyofAging,22,932-930.Pradhan,N.,andDutt,D.N.993.AnonlinearperspectiveinunderstandingtheneurodynamicsofEEG.ComputersinBiologyandMedicine,23,425-442.Pradhan,N.,andSadasivan,P.K.96.ThenatureofdominantLyapunovex-ponentandattractordimensioncurvesofEEGinsleep.ComputersinBiologyandMedicine,26,419-428.Pratico,D.,Uryu,K.,Leight,S.,Trojanowski,J.K.,andLee,V.M.-Y.001.IncreasedlipidperoxidationprecedesamyloidplaqueformationinananimalmodelofAlzheimeramyloidosis.JournalofNeuroscience,21,4183-4187.Price,D.L.,andSisodia,S.S.1998.MutantgenesinfamilialAlzheimer'sdiseaseandtransgenicmodels.AnnualReviewofNeuroscience,21,479-505.Price,J.,Ko,A.,Wade,M.,Tsou,S.,McKeel,D.andMorris,J.001.NeuronnumberintheentorhinalcortexandCA1inpreclinicalAlzheimerdisease.ArchivesofNeurology,58,1395-1402.Price,R.K.,Spitznagel,E.L.,Downey,T.J.,Meyer,D.J.,andRisk,N.K.000.Applyingarticialneuralnetworkmodelstoclinicaldecisionmaking.PsychologicalAssessment,12,40-51.Pritchard,W.S.,andDuke,D.W.95.Measuringchaosinthebrain:AtutorialreviewofEEGdimensionestimation.BrainandCognition,27,353-397.Provost,F.,andFawcett,T.001.Robustclassicationforimpreciseenviron-ments.MachineLearning,42,203-231.Prusiner,S.B.91.Molecularbiologyofpriondiseases.Science,252,1515-1522.Pugh,P.L.,Ahmed,S.F.,Smith,M.I.,Upton,N.,andHunter,A.J.004.AbehavioralcharacterizationoftheFVB/Nmousestrain.BehavioralBrainResearch,155,283-289.Pugh,P.L.,Richardson,J.C.,Bate,S.T.,Upton,N.,andSunter,D.007.Non-cognitivebehaviorsinanAPP/PS1transgenicmodelofAlzheimer'sdisease.BehavioralBrainResearch,178,18-28.Qizilbash,N.,Whitehead,A.,Higgins,J.,Wilcock,G.,Schneider,L.,andFarlow,M.998.CholinesteraseinhibitionforAlzheimerdisease:Ameta-analysisofthetacrinetrials.JournaloftheAmericanMedicalAssociation,280,1777-1782.293

PAGE 304

Quayhagen,M.andQuayhagen,M.89.Dierentialeectsoffamily-basedstrategiesonAlzheimer'sdisease.TheGerontologist,29,150-155.Quinlan,J.R.986.Inductionofdecisiontrees.MachineLearning,1,81-106.Quinlan,J.R.93.C4.5:ProgramsforMachineLearning.SanMateo,CA:MorganKaufmannPublishers.Ramirez,B.G.,Blazquez,C.,delPulgar,T.G.,Guzman,M.,anddeCeballos,M.L.2005.PreventionofAlzheimer'sdiseasepathologybycannabinoids:Neuroprotectionmediatedbyblockadeofmicroglialactivation.JournalofNeuroscience,25,1904-1913.RammesG,RupprechtR,FerrariU,ZieglgansbergerW,andParsonsCG.001.TheN-methyl-D-aspartatereceptorchannelblockersmemantine,MRZ2/579andotheramino-alkyl-cyclohexanesantagonise5-HTreceptorcurrentsinculturedHEK-293andN1E-115cellsystemsinanon-competitivemanner.NeuroscienceLetters,306,81-4.Ramos,A.,andMormede,P.1998.Stressandemotionality:Amultidimensionalandgeneticapproach.NeuroscienceandBiobehavioralReviews,22,33-57.Raskind,M.,Sadowsky,C.,Sigmund,W.,Beitler,P.andAuster,S.997.Eectoftacrineonlanguage,praxis,andnoncognitivebehavioralproblemsinAlzheimerdisease.ArchivesofNeurology,54,836-840.Ratcli,R.990.Connectionistmodelsofrecognitionmemory:Constraintsim-posedbylearningandforgettingfunctions.PsychologicalReview,97,285-308.Rauch,S.L.,Whalen,P.J.,Savage,C.R.,Curran,T.,Kendrick,A.,Brown,H.D.,Bush,G.,Breiter,H.C.,andRosen,B.R.997.Striatalrecruitmentduringanimplicitsequencelearningtaskasmeasuredbyfunctionalmagneticresonanceimaging.HumanBrainMapping,5,124-132.Raudys,S.J.,andJain,A.K.91.Smallsamplesizeeectsinstatisticalpatternrecognition:Recommendationsforpractitioners.IEEETransactionsonPatternRecognitionandMachineIntelligence,13,252-264.Ray,S.,Britschgi,M.,Herbert,C.,Takeda-Uchimura,Y.,Boxer,A.,Blennow,K.,Friedman,L.F.,Galasko,D.R.,Jutel,M.,Karydas,A.,Kaye,J.A.,Leszek,J.,Miller,B.L.,Minthon,L.,Quinn,J.F.,Rabinovici,G.D.,Robinson,W.H.,Sabbagh,M.N.,So,Y.T.,Sparks,D.L.,Tabaton,M.,Tinklenberg,J.,Yesavage,J.A.,Tibshirani,R.,andWyss-Coray,T.007.ClassicationandpredictionofclinicalAlzheimer'sdiagnosisbasedonplasmasignalingproteins.NatureMedicine,13,1359-1362.294

PAGE 305

Reber,A.S.67.Implicitlearningofarticialgrammars.JournalofVerbalLearningandVerbalBehavior,6,855-863.Rechtschaen,A.,andKales,A.968.AManualofStandardizedTerminology,TechniquesandScoringSystemforSleepStagesinHumanSubjects.NIHPub-licationNo.204.Washington,DC:PublicHealthServices,U.S.GovernmentPrintingOce.Reddy,P.H.006.Amyloidprecursorprotein-mediatedfreeradicalsandoxida-tivedamage:ImplicationsforthedevelopmentandprogressionofAlzheimer'sdisease.JournalofNeurochemistry,96,1-13.Reed,J.,andJohnson,P.994.Assessingimplicitlearningwithindirecttests:Determiningwhatislearnedaboutsequencestructure.JournalofExperimen-talPsychology:Learning,Memory,andCognition,20,585-594.Reilly,J.F.,Games,D.,Rydel,R.E.,Freedman,S.,Schenk,D.,Young,W.G.,Morrison,J.H.,andBloom,F.E.03.Amyloiddepositioninthehip-pocampusandentorhinalcortex:Quantitativeanalysisofatransgenicmousemodel.ProceedingsoftheNationalAcademyofSciencesUSA,100,4837-4842.Reiman,E.M.,Webster,J.A.,Myers,A.J.,Hardy,J.,Dunckley,T.,Zismann,V.L.,Joshipura,K.D.,Pearson,J.V.,Hu-Lince,D.,Huentelman,M.J.,Craig,D.W.,Coon,K.D.,Liang,W.S.,Herbert,R.H.,Beach,T.,Rohrer,K.C.,Zhao,A.S.,Leung,D.,Bryden,L.,Marlowe,L.,Kaleem,M.,Mastroeni,D.,Grover,A.,Heward,C.B.,Ravid,R.,Rogers,J.,Hutton,M.L.,Melquist,S.,Petersen,R.C.,Alexander,G.E.,Caselli,R.J.,Kukull,W.,Papassotiropou-los,A.,andStephan,D.A.07.GAB2allelesmodifyAlzheimer'sriskinAPOEepsilon4carriers.Neuron,54,713-720.Reisberg,B.,Doody,R.,Stoer,A.,Schmitt,F.,Ferris,S.,andMobius,H.J.003.Memantineinmoderate-to-severeAlzheimer'sdisease.NewEnglandJournalofMedicine,348,1333-1341.Reisberg,B.,Franssen,E.H.,Hasan,S.M.,Monteiro,I.,Boksay,I.,Souren,L.E.,Kenowsky,S.,Auer,S.R.,Elahi,S.,andKluger,A.999.Retrogenesis:Clinical,physiologic,andpathologicmechanismsinbrainaging,Alzheimer'sandotherdementingprocesses.EuropeanArchivesofPsychiatryandClinicalNeurosciences,249Suppl3,28-36.Riek,R.006.InfectiousAlzheimer'sdisease?Nature,444,429-431.Risner,M.E.,Saunders,A.M.,Altman,J.F.B.,Ormandy,G.C.,Craft,S.,Foley,I.M.,Zvartau-Hind,M.E.,Hosford,D.A.,andRoses,A.D.06.E-cacyofrosiglitazoneinageneticallydenedpopulationwithmild-to-moderateAlzheimer'sdisease.ThePharmacogenomicsJournal,6,246-254.295

PAGE 306

Ritchie,K.,Carriere,I.,deMendonca,A.,Portet,F.,Dartigues,J.F.,Rouaud,O.,Barberger-Gateau,P.,andAncelin,M.L.07.Theneuroprotectiveeectsofcaeine.AprospectivepopulationstudytheThreeCityStudy.Neurology,69,536-545.Roberson,E.D.,Scearce-Levie,K.,Palop,J.J.,Yan,F.,Cheng,I.H.,Wu,T.,Gerstein,H.,Yu,G.Q.,andMucke,L.07.Reducingendogenoustauame-lioratesamyloidbeta-induceddecitsinanAlzheimer'sdiseasemousemodel.Science,316,750-754.Robert,C.,Guilpin,C.,andLimoge,A.97.Comparisonbetweenconventionalandneuralnetworkclassiersforratsleep-wakestagediscrimination.Neu-ropsychobiology,35,221-225.Robert,C.,Karasinski,P.,Natowicz,R.,andLimoge,A.96.AdultratvigilancestatesdiscriminationbyarticialneuralnetworksusingasingleEEGchannel.PhysiologyandBehavior,59,1051-1060.Roberts,W.A.,andDale,R.H.I.81.Remembranceofplaceslasts:Proac-tiveinterferenceandpatternsofchoiceinratspatialmemory.LearningandMotivation,12,261-281.Robinson,D.M.,andKeating,G.M.2006.Memantine:AreviewofitsuseinAlzheimer'sdisease.Drugs,66,1515-1534.Rocchi,A.,Pellegrini,S.,Siciliano,G.,andMurri,L.03.Causativeandsus-ceptibilitygenesforAlzheimer'sdisease:Areview.BrainResearchBulletin,61,1-24.Rockenstein,E.M.,McConlogue,L.,Tan,H.,Power,M.,Masliah,E.,andMucke,L.95.Levelsandalternativesplicingofamyloidbetaproteinprecur-sorAPPtranscriptsinbrainsofAPPtransgenicmiceandhumanswithAlzheimer'sdisease.JournalofBiologicalChemistry,270,28257-28267.Rodgers,R.J.,andJohnson,N.J.T.95.Factoranalysisofspatiotemporalandethologicalmeasuresinthemurineelevatedplus-mazetestofanxiety.PharmacologyBiochemistryandBehavior,52,297-303.Roertgen,K.E.,Parisi,J.E.,Clark,H.B.,Barnes,D.L.,O'Brien,T.D.,andJohnson,K.H.96.A-associatedcerebralangiopathyandsenileplaqueswithneurobrillarytanglesandcerebralhemorrhageinanagedwolverineGulogulo.NeurobiologyofAging,17,243-247.296

PAGE 307

Rogaev,E.I.,Sherrington,R.,Rogaeva,E.A.,Levesque,G.,Ikeda,M.,Liang,Y.,Chi,H.,Lin,C.,Holman,K.,Tsuda,T.,Mar,L.,Sorbi,S.,Nacmias,B.,Placentini,S.,Amaducci,L.,Chumakov,I.,Cohen,D.,Lannfelt,L.,Fraser,P.E.,Rommens,J.M.,andStGeorge-Hyslop,P.H.95.FamilialAlzheimer'sdiseaseinkindredswithmissensemutationsinageneonchromosome1relatedtotheAlzheimer'sdiseasetype3gene.Nature,376,775-778.Rogaev,E.I.,Sherrington,R.,Wu,C.,Levesque,G.,Liang,Y.,Rogaeva,E.A.,Ikeda,M.,Holman,K.,Lin,C.,Lukiw,W.J.,deJong,P.J.,Fraser,P.E.,Rommens,J.M.,St.George-Hyslop,P.1997,Analysisofthe5-primesequence,genomicstructure,andalternativesplicingofthepresenilin-1genePSEN1associatedwithearlyonsetAlzheimerdisease.Genomics,40,415-424.Rogaeva,E.,Meng,Y.,Lee,J.H.,Gu,Y.,Kawarai,T.,Zou,F.,Katayama,T.,Baldwin,C.T.,Cheng,R.,Hasegawa,H.,Chen,F.,Shibata,N.,Lunetta,K.L.,Pardossi-Piquard,R.,Bohm,C.,Wakutani,Y.,Cupples,L.A.,Cuenco,K.T.,Green,R.C.,Pinessi,L.,Rainero,I.,Sorbi,S.,Bruni,A.,Duara,R.,Friedland,R.P.,Inzelberg,R.,Hampe,W.,Bujo,H.,Song,Y.-Q.,Andersen,O.M.,Willnow,T.E.,Gra-Radford,N.,Petersen,R.C.,Dickson,D.,Der,S.D.,Fraser,P.E.,Schmitt-Ulms,G.,Younkin,S.,Mayeux,R.,Farrer,L.A.andSt.George-Hyslop,P.007.Theneuronalsortilin-relatedreceptorSORL1isgeneticallyassociatedwithAlzheimerdisease.NatureGenetics,39,168-177.Rogers,D.C.,Fisher,E.M.C.,Brown,S.D.M.,Peters,J.,Hunter,A.J.,andMartin,J.E.997.Behavioralandfunctionalanalysisofmousephenotype:SHIRPA,aproposedprotocolforcomprehensivephenotypeassessment.Mam-malianGenome,8,711-713.Rogers,D.C.,Jones,D.N.C.,Nelson,P.R.,Jones,C.M.,Quilter,C.A.,Robinson,T.L.,andHagan,J.J.999.UseofSHIRPAanddiscriminantanalysistocharacterizemarkeddierencesinthebehavioralphenotypeofsixinbredmousestrains.BehavioralBrainResearch,105,207-217.Rogers,S.L.,Farlow,M.R.,Doody,R.S.,Mohs,R.,andFriedho,L.T.998.A24-week,double-blind,placebo-controlledtrialofdonepezilinpatientswithAlzheimer'sdisease.Neurology,50,136-145.Roher,A.,Baudry,J.,Chaney,M.,Kuo,Y.,Stine,W.andEmmerling,M.000.Oligomerizationandbrilassemblyoftheamyloid-bprotein.BiochimicaetBiophysicaActa,1502,31-43.297

PAGE 308

Roher,A.,Wolfe,D.,Palutke,M.,andKuKuruga,D.86.Purication,ul-trastructure,andchemicalanalysisofAlzheimerdiseaseamyloidplaquecoreprotein.ProceedingsoftheNationalAcademyofSciencesUSA,83,2662-2666.Rolland,Y.,Pillard,F.,Klapouszczak,A.,Reynish,E.,Thomas,D.,Andrieu,S.,Riviere,D.,andVellas,B.2007.ExerciseprogramfornursinghomeresidentswithAlzheimer'sdisease:A1-yearrandomized,controlledtrial.JournaloftheAmericanGeriatricsSociety,55,158-165.Rosen,R.F.,Farberg,A.S.,Gearing,M.,Dooyema,J.,Long,P.M.,Anderson,D.C.,Davis-Turak,J.,Coppola,G.,Geschwind,D.H.,Pare,J.-F.,Duong,T.Q.,Hopkins,W.D.,Preuss,T.M.,andWalker,L.C.008.Tauopathywithpairedhelicallamentsinanagedchimpanzee.JournalofComparativeNeurology,509,259-270.Rosenblatt,F.958.Theperceptron:Aprobabilisticmodelforinformationstor-ageandorganizationinthebrain.PsychologicalReview,65,386-408.Rossler,M.,Zarski,R.,Bohl,J.andOhm,T.02.Stage-dependentandsector-specicneuronallossinhippocampusduringAlzheimer'sdisease.ActaNeu-ropathologica,103,363-369.Rosso,A.,Mossey,J.,andLippa,C.F.008.Caeine:NeuroprotectivefunctionsincognitionandAlzheimer'sdisease.AmericanJournalofAlzheimer'sDiseaseandOtherDementias,23,417-422.Rousseau,J.B.,VanLochem,P.B.,Gispen,W.H.,andSpruijt,B.M.000.Classicationofratbehaviorwithanimage-processingmethodandaneuralnetwork.BehavioralResearchMethods,Instrumentation,andComputers,32,63-71.Rovio,S.,Kareholt,I.,Viitanen,M.,Winblad,B.,Tuomilehto,J.,Soininen,H.,Nissinen,A.,andKivipelto,M.07.Work-relatedphysicalactivityandtheriskofdementiaandAlzheimer'sdisease.InternationalJournalofGeriatricPsychiatry,22,874-882.Rowan,A.J.,andTolunsky,E.003.PrimerofEEG.Philadelphia,PA:Elsevier.Ruehl,W.W.,Bruyette,D.S.,DePaoli,A.,Cotman,C.W.,Head,E.,Milgram,N.W.,andCummings,B.J.95.Caninecognitivedysfunctionasamodelforhumanagerelatedcognitivedecline,dementiaandAlzheimer'sdisease:Clinicalpresentation,cognitivetesting,pathologyandresponsetoL-deprenyltherapy.ProgressinBrainResearch,106,217-225.Rumelhart,D.E.,Hinton,G.E.,andWilliams,R.J.986.Learningrepresenta-tionsbyback-propagatingerrors.Nature,323,533-536.298

PAGE 309

Rumelhart,D.E.,andMcClelland,J.L.1986.ParallelDistributedProcessing:ExplorationsintheMicrostructureofCognition.Cambridge,MA:MITPress.Salmon,D.P.,Thomas,R.G.,Pay,M.M.,Booth,A.,Hofstetter,C.R.,Thal,L.J.,etal.02.Alzheimer'sdiseasecanbeaccuratelydiagnosedinverymildlyimpairedindividuals.Neurology,59,1022-1028.Sanan,D.,Weisgraber,K.,Russell,S.,Mahley,R.,Huang,D.,Saunders,A.,Schmechel,D.,Wisniewski,T.,Frangione,B.,Roses,A.andStrittmatter.994.ApolipoproteinEassociateswithbetaamyloidpeptideofAlzheimer'sdiseasetoformnovelmonobrils.IsoformapoE4associatesmoreecientlythanapoE3.JournalofClinicalInvestigation,94,860-869.Sango,K.,McDonald,M.P.,Crawley,J.N.,Mack,M.L.,Tit,C.J.,Skop,E.,Starr,C.M.,Homann,A.,Sandho,K.,Suzuki,K.,andProia,R.L.996.Micelackingbothsubunitsoflysosomalbeta-hexosaminidasedisplaygangliosidosisandmucopolysaccharidosis.NatureGenetics,14,348-352.Savonenko,A.,Xu,G.M.,Melnikova,T.,Morton,J.L.,Gonzales,V.,Wong,M.P.F.,Price,D.L.,Tang,F.,Markowska,A.L.,andBorchelt,D.R.005.Episodic-likememorydecitsintheAPPswe/PS1dE9mousemodelofAlzheimer'sdisease:Relationshipsto-amyloiddepositionandneurotrans-mitterabnormalities.NeurobiologyofDisease,18,602-617.Scarmeas,N.,Albert,S.M.,Manly,J.J.,andStern,Y.006a.EducationandratesofcognitivedeclineinincidentAlzheimer'sdisease.JournalofNeurology,Neurosurgery,andPsychiatry,77,308-316.Scarmeas,N.,Stern,Y.,Mayeux,R.,andLuschsinger,J.2006b.Mediterraneandiet,Alzheimer'sdisease,andvascularmediation.ArchivesofNeurology,63,1709-1717.Scarpini,E.,Scheltens,P.andFeldman,H.003.TreatmentofAlzheimer'sdis-ease:Currentstatusandnewprospectives.LancetNeurology,2,539-47.Schacter,D.L.87.Implicitmemory:Historyandcurrentstatus.JournalofExperimentalPsychology:Learning,Memory,andCognition,13,501-518.Scherder,E.,Eggermont,L.,Sergeant,J.,andBoersma,F.07.PhysicalactivityandcognitioninAlzheimer'sdisease:Relationshiptovascularriskfactors,executivefunctionsandgait.ReviewsofNeuroscience,18,149-158.Schollhorn,W.I.2004.Applicationsofarticialneuralnetsinclinicalbiome-chanics.ClinicalBiomechanics,19,876-898.299

PAGE 310

Schlosser,R.,Hutchinson,M.,Joseer,S.,Rusinek,H.,Saarimaki,A.,Stevenson,J.,Dewey,S.L.,andBrodie,J.D.98.Functionalmagneticresonanceimagingofhumanbrainactivityinaverbaluencytask.JournalofNeurology,NeurosurgeryandPsychiatry,64,492-498.Schwab,C.,Hosokawa,M.,andMcGeer,P.L.04.Transgenicmiceoverex-pressingamyloidbetaproteinareanincompletemodelofAlzheimer'sdisease.ExperimentalNeurology,188,52-64.Scott,A.,andWild,C.91.Transformationsandr-squared.AmericanStatisti-cian,45,127-128.Scott,L.J.,andGoa,K.L.000.Galantamine:AreviewofitsuseinAlzheimer'sdisease.Drugs,60,1095-1122.Seaman,J.W.,Walls,S.C.,Wide,S.E.,andJaeger,R.G.94.Caveatemptor:Ranktransformmethodsandinteractions.TrendsinEcologyandEvolution,9,261-263.Selkoe,D.J.89.BiochemistryofalteredbrainproteinsinAlzheimer'sdisease.AnnualReviewofNeuroscience,12,463-490.Selkoe,D.J.001.Alzheimer'sdisease:Genes,proteins,andtherapy.Physiolog-icalReviews,81,741-766.Selkoe,D.J.,Bell,D.S.,Podlisny,M.B.,Price,D.L.,andCork,L.C.987.ConservationofbrainamyloidproteinsinagedmammalsandhumanswithAlzheimer'sdisease.Science,235,873-877.Selkoe,D.J.,andPodlisny,M.B.002.DecipheringthegeneticbasisofAlzheimer'sdisease.AnnualReviewofGenomicsandHumanGenetics,3,67-99.Serpell,L.000.Alzheimer'samyloidbrils:structureandassembly.BiochimicaetBiophysicaActa,1502,16-30.Serretti,A.,Zanardi,R.,Mandelli,L.,Smeraldi,E.,andColombo,C.007.Aneuralnetworkmodelforcombiningclinicalpredictorsofantidepressantresponseinmooddisorders.JournalofAectiveDisorders,98,239-245.Sevush,S.,Guterman,A.,andVillalon,A.V.91.Improvedverballearn-ingafteroutpatientphysostigminetherapyinpatientswithdementiaoftheAlzheimer'stype.JournalofClinicalPsychiatry,52,300-303.Sevush,S.,Peruyera,G.,Bertran,A.,andCisneros,W.03.Athree-factormodelofcognitioninAlzheimerdisease.CognitiveandBehavioralNeurology,16,110-117.300

PAGE 311

Shallice,T.,andWarrington,E.K.970.Independentfunctioningofverbalmemorystores:Aneurophysiologicalstudy.QuarterlyJournalofExperimentalPsychology,22,261-273.Shenk,D.001.TheForgetting.Alzheimers:PortraitofanEpidemic.NewYork:Anchor.Sherrington,R.,Rogaev,E.I.,Liang,Y.,Rogaeva,E.A.,Levesque,G.,Ikeda,M.,Chi,H.,Lin,C.,Li,G.,Holman,K.,Tsuda,T.,Mar,L.,Foncin,J.F.,Bruni,A.C.,Montesi,M.P.,Sorbi,S.,Rainero,I.,Pinessi,L.,Nee,L.,Chumakov,I.,Pollen,D.,Brookes,A.,Sanseau,P.,Polinsky,R.J.,Wasco,W.,DaSilva,H.A.R.,Haines,J.L.,Pericak-Vance,M.A.,Tanzi,R.E.,Roses,A.D.,Fraser,P.E.,Rommens,J.M.,StGeorge-Hyslop,P.H.95,Cloningofagenebearingmis-sensemutationsinearly-onsetfamilialAlzheimer'sdisease.Nature,375,754-760.Shif,O.,Gillette,K.,Damkaoutis,C.M.,Carrano,C.,Robbins,S.J.,andHoman,J.R.06.EectsofGinkgobilobaadministeredafterspatiallearningonwa-termazeandradialarmmazeperformanceinyoungadultrats.Pharmacology,BiochemistryandBehavior,84,17-25.Shivers,B.D.,Hilbich,C.,Multhaup,G.,Salbaum,M.,Beyreuther,K.,andSee-burg,P.H.988.Alzheimer'sdiseaseamyloidogenicglycoprotein:Expres-sionpatterninratbrainsuggestsaroleincellcontact.EMBOJournal,7,1365-1370.Shore,D.I.,Stanford,L.,MacInnes,W.J.,Klein,R.M.,andBrown,R.E.001.Ofmiceandmen:VirtualHebb-Williamsmazespermitcomparisonofspatiallearningacrossspecies.Cognitive,Aective,andBehavioralNeuroscience,1,83-89.Shortlie,E.H.74.MYCIN:Arule-basedcomputerprogramforadvisingphysi-ciansregardingantimicrobialtherapyselection.Doctoraldissertation,StanfordUniversity.Shortlie,E.H.76.Computer-BasedMedicalConsultations:MYCIN.NewYork:AmericanElsevier.Sigut,J.,Pineiro,J.,Gonzalez,E.,andTorres,J.07.Anexpertsystemforsupervisedclassierdesign:ApplicationtoAlzheimer'sdiagnosis.ExpertSys-temswithApplications,32,927-938.Silvestrelli,G.,Lanari,A.,Parnetti,L.,Tomassoni,D.,andAmenta,F.006.TreatmentofAlzheimer'sdisease:Frompharmacologytoabetterunderstand-ingofdiseasepathophysiology.MechanismsofAgeingandDevelopment,127,148-157.301

PAGE 312

Simard,A.R.,Soulet,D.,Gowing,G.,Julien,J.P.,andRivest,S.006.Bonemarrow-derivedmicrogliaplayacriticalroleinrestrictingsenileplaqueforma-tioninAlzheimer'sdisease.Neuron,49,489-502.Sitzer,D.I.,Twamley,E.W.,andJeste,D.V.006.CognitivetraininginAlzheimer'sdisease:Ameta-analysisoftheliterature.ActaPsychiatricaScan-dinavica,114,75-90.Sizova,D.,Charbaut,E.,Delalande,F.,Poirier,F.,High,A.A.,Parker,F.,VanDorsselaer,A.,Duchesne,M.,andDiu-Hercend,A.007.ProteomicanalysisofbraintissuefromanAlzheimer'sdiseasemousemodelbytwo-dimensionaldierencegelelectrophoresis.NeurobiologyofAging,28,357-370.Skarda,C.A.,andFreeman,W.J.987.Howbrainsmakechaosinordertomakesenseoftheworld.BehavioralBrainScience,10,161-195.Smith,A.002.Eectsofcaeineonhumanbehavior.FoodandChemicalToxi-cology,40,1243-1255.Sobow,T.,Flirski,M.,andLiberski,P.P.004.Amyloid-betaandtauproteinsasbiochemicalmarkersofAlzheimer'sdisease.ActaNeurobiologiaeExperi-mentalis,64,53-70.Song,J.,Burrage,K.,Yuan,Z.,andHuber,T.006.Predictionofcis/transisomerizationinproteinsusingPSI-BLASTprolesandsecondarystructureinformation.BMCBioinformatics,7,124.Souren,L.,Franssen,E.andReisberg,B.95.ContracturesandlossoffunctioninpatientswithAlzheimer'sdisease.JournaloftheAmericanGeriatricSociety,43,650-655.Spalletta,G.,Baldinetti,F.,Buccione,I.,Fadda,L.,Perri,R.,Scalmana,S.,Serra,L.,andCaltagirone,C.004.CognitionandbehaviorareindependentandheterogeneousdimensionsinAlzheimer'sdisease.JournalofNeurology,251,688-695.Spear,N.E.,Miller,J.S.,andJagielo,J.A.90.Animalmemoryandlearning.AnnualReviewsinPsychology,41,169-211.Spearman,C.1904.Generalintelligence"objectivelydeterminedandmeasured.AmericanJournalofPsychology,15,201-293.Sperling,G.1960.Theinformationavailableinbriefvisualpresentation.Psycho-logicalMonographs,74,1-29.Squire,L.R.992.Memoryandthehippocampus:Asynthesisfromndingswithrats,monkeys,andhumans.PsychologicalReview,99,195-231.302

PAGE 313

Squire,L.R.,andZola,S.M.97.Amnesia,memoryandbrainsystems.Philo-sophicalTransactionsoftheRoyalSocietyofLondon,SeriesB,352,1663-1673.Stackman,R.W.,Eckenstein,F.,Frei,B.,Kulhanek,D.,Nowlin,J.,andQuinn,J.F.2003.Preventionofage-relatedspatialmemorydecitsinatransgenicmousemodelofAlzheimer'sdiseasebychronicGinkgobilobatreatment.Ex-perimentalNeurology,184,510-520.Stein,T.D.,andJohnson,J.A.002.Lackofneurodegenerationintransgenicmiceoverexpressingmutantamyloidprecursorproteinisassociatedwithin-creasedlevelsoftransthyretinandtheactivationofcellsurvivalpathways.TheJournalofNeuroscience,22,7380-7388.Steinberger,D.,Reynolds,D.S.,Ferris,P.,Lincoln,R.,Datta,S.,Stanley,J.,Paterson,A.,Dawson,G.R.,andFlint,J.003.Geneticmappingofvariationinspatiallearninginthemouse.JournalofNeuroscience,23,2426-2433.Stelzmann,R.A.,Schnitzlein,H.N.,andMurtagh,F.R.995.AnEnglishtranslationofAlzheimer's1907paper,UbereineeigenartigeErkrankungderHirnrinde."ClinicalAnatomy,8,429-431.Stern,Y.,Gurland,B.,Tatemichi,T.K.,Tang,M.X.,Wilder,D.,andMayeux,R.994.InuenceofeducationandoccupationontheincidenceofAlzheimer'sdisease.JournaloftheAmericanMedicalAssociation,27113,1004-1010.Stern,E.A.,Bacskai,B.J.,Hickey,G.A.,Attenello,F.J.,Lombardo,J.A.,andHyman,B.T.2004.Corticalsynapticintegrationinvivoisdisruptedbyamyloid-betaplaques.JournalofNeuroscience,24,4535-4540.Stewart,W.F.,Kawas,C.,Corrada,M.,andMetter,E.J.97.RiskofAlzheimer'sdiseaseanddurationofNSAIDuse.Neurology,48,626632.StGeorge-Hyslop,P.H.,Tanzi,R.E.,Polinsky,R.J.,Haines,J.L.,Nee,L.,Watkins,P.C.,Myers,R.H.,Feldman,R.G.,Pollen,D.,Drachman,D.,etal.1987.ThegeneticdefectcausingfamilialAlzheimer'sdiseasemapsonchromosome21.Science,235,885-890.Stone,M.77.Asymptoticsforandagainstcross-validation.Biometrika,64,29-35.Stopford,C.L.,Snowden,J.S.,Thompson,J.C.,andNeary,D.007.DistinctmemoryprolesinAlzheimer'sdisease.Cortex,43,846-857.Strauss,M.E.,andFritsch,T.004.FactorstructureoftheCERADneuropsy-chologicalbattery.JournaloftheInternationalNeuropsychologicalSociety,10,559-565.303

PAGE 314

Stricker,L.J.,andRock,D.A.1987.FactorstructureoftheGREGeneralTestinyoungandmiddleadulthood.DevelopmentalPsychology,23,526-536.Stroop,J.R.35.Studiesofinterferenceinserialverbalreactions.JournalofExperimentalPsychology,18,643-661.Stroustrup,B.85.TheC++ProgrammingLanguage.NewYork:Addison-Wesley.Sturchler-Pierrat,C.,Abramowski,D.,Duke,M.,Wiederhold,K.,Mistl,C.,Roth-acher,S.,Ledermann,B.,Burki,K.,Frey,P.,Paganetti,P.,Waridel,C.,Cal-houn,M.,Jucker,M.,Probst,A.,Staufenbiel,M.,andSommer,B.997.TwoamyloidprecursorproteintransgenicmousemodelswithAlzheimerdisease-likepathology.ProceedingsoftheNationalAcademyofSciencesUSA,94,13287-13292.Suh,Y.-H.,andChecler,F.02.Amyloidprecursorprotein,presenilins,andalpha-synuclein:MolecularpathogenesisandpharmacologicalapplicationsinAlzheimer'sdisease.PharmacologicalReviews,54,469-525.Suo,Z.,Cox,A.A.,Bartelli,N.,Rasul,I.,Festo,B.W.,Premont,R.T.,andAren-dash,G.W.07.GRK5deciencyleadstoearlyAlzheimer-likepathologyandworkingmemoryimpairment.NeurobiologyofAging,28,1873-1888.Suo,Z.,Wu,M.,Citron,B.A.,Wong,G.T.,andFesto,B.W.004.Ab-normalityofG-protein-coupledreceptorkinasesatprodromalandearlystagesofAlzheimer'sdisease:Anassociationwithearlybeta-amyloidaccumulation.JournalofNeuroscience,24,3444-3452.Suzuki,S.,Augerinos,G.,andBlack,A.H.980.Stimuluscontrolofspatialbehaviorontheeight-armmazeinrats.LearningandMotivation,11,1-18.Sweatt,J.1999.Towardamolecularexplanationforlong-termpotentiation.LearningandMemory,6,399-416.Swets,J.A.988.Measuringtheaccuracyofdiagnosticsystems.Science,240,1285-1293.Swets,J.A.,Dawes,R.M.,andMonahan,J.2000.Betterdecisionsthroughscience.ScienticAmerican,2834,82-87.Tabachnick,B.G.,andFidell,L.S.01.UsingMultivariateStatistics.NewYork:HarperCollins.304

PAGE 315

Taddei,K.,Fisher,C.,Laws,S.M.,Martins,G.,Paton,A.,Clarnette,R.M.,Chung,C.,Brooks,W.S.,Hallmayer,J.,Miklossy,J.,Relkin,N.,StGeorge-Hyslop,P.H.,Gandy,S.E.,andMartins,R.N.02.Associationbetweenpresenilin-1Glu318GlymutationandfamilialAlzheimer'sdiseaseintheAus-tralianpopulation.MolecularPsychiatry,7,776-781.Tafeit,E.,andReibnegger,G.99.Articialneuralnetworksinlaboratorymedicineandmedicaloutcomeprediction.ClinicalChemistryandLaboratoryMedicine,37,845-853.Takeuchi,A.,Irizarry,M.,Du,K.,Saido,T.,HsiaoAshe,K.,Hasegawa,M.,Mann,D.,Hyman,B.,andIwatsubo,T.000.Age-relatedamyloiddepositionintransgenicmiceoverexpressingbothAlzheimermutantpresenilin1andamyloidprecursorproteinSwedishmutantisnotassociatedwithglobalneuronalloss.AmericanJournalofPathology,157,331-339.Tandon,R.,Adak,S.,andKaye,J.A.06.NeuralnetworksforlongitudinalstudiesinAlzheimer'sdisease.ArticialIntelligenceinMedicine,36,245-255.Tanzi,R.E.,andBertram,L.01.NewfrontiersinAlzheimer'sdiseasegenetics.Neuron,32,181-184.Tanzi,R.E.,andBertram,L.005.TwentyyearsoftheAlzheimer'sdiseaseamyloidhypothesis:Ageneticperspective.Cell,120,545-555.Tanzi,R.E.,Gusella,J.F.,Watkins,P.C.,Bruns,G.A.,StGeorge-Hyslop,P.,VanKeuren,M.L.,Patterson,D.,Pagan,S.,Kurnit,D.M.,andNeve,R.L.987.Amyloidbetaproteingene:cDNA,mRNAdistribution,andgeneticlinkageneartheAlzheimerlocus.Science,235,880-884.Tanzi,R.E.,McClatchey,A.I.,Lampert,E.D.,Villa-Komaro,L.,Gusella,J.F.,andNeve,R.L.1988.ProteaseinhibitordomainencodedbyanamyloidproteinprecursormRNAassociatedwithAlzheimer'sdisease.Nature,331,528530.Tariot,P.,Solomon,P.,Morris,J.,Kershaw,P.,Lilienfeld,S.andDing,C.000.A5-month,randomized,placebo-controlledtrialofgalantamineinAD.TheGalantamineUSA-10StudyGroup.Neurology,54,2269-2276.Tariot,P.N.,Farlow,M.R.,Grossberg,G.T.,Graham,S.M.,McDonald,S.,andGergel,I.04.MemantinetreatmentinpatientswithmoderatetosevereAlzheimer'sdiseasealreadyreceivingdonezepil:Arandomizedcontrolledtrial.JournaloftheAmericanMedicalAssociation,291,317-324.Tecott,L.H.,andNestler,E.J.004.Neurobehavioralassessmentintheinfor-mationage.NatureNeuroscience,7,462-466.305

PAGE 316

Tekirian,T.L.,Cole,G.M.,Russell,M.J.,Yang,F.,Wekstein,D.R.,Patel,E.,Snowdon,D.A.,Markesbery,W.R.,andGeddes,J.W.96.Carboxyterminalof-amyloiddepositsinagedhuman,canine,andpolarbearbrains.NeurobiologyofAging,17,249-257.Thompson,P.,Hayashi,K.,deZubicaray,G.,Janke,A.,Rose,S.,Semple,J.,Herman,D.,Hong,M.,Dittmer,S.,Doddrell,D.andToga,A.003.Dy-namicsofgraymatterlossinAlzheimer'sdisease.JournalofNeuroscience,23,994-1005.Thompson,R.F.,andKim,J.J.96.Memorysystemsinthebrainandlocal-izationofamemory.ProceedingsoftheNationalAcademyofSciencesUSA,93,13438-13444.Thorndike,R.L.935.Organizationofbehaviorinthealbinorat.GeneticPsychologyMonographs,17,1-70.Thurstone,L.L.38.Primarymentalabilities.PsychometricMonographs,1.Tian,H.,andShang,Z.06.Articialneuralnetworkasaclassicationmethodofmicebytheircalls.Ultrasonics,44,e275-e278.Toer,A.970.FutureShock.NewYork:RandomHouse.Toga,A.W.,andThompson,P.M.005.Geneticsofbrainstructureandintelli-gence.AnnualReviewofNeuroscience,28,1-23.Tomidokoro,Y.,Harigaya,Y.,Matsubara,E.,Ikeda,M.,Kawarabayashi,T.,Shi-rao,T.,Ishiguro,K.,Okamoto,K.,Younkin,S.G.,andShoji,M.2001.BrainAbetaamyloidosisinAPPswmiceinducesaccumulationofpresenilin-1andtau.JournalofPathology,194,500-506.Tomita,T.,Maruyama,K.,Saido,T.C.,Kume,H.,Shinozaki,K.,Tokuhiro,S.,Capell,A.,Walter,J.,Grunberg,J.,Haass,C.,Iwatsubo,T.,andObata,K.997.Thepresenilin2mutationN141IlinkedtofamilialAlzheimerdiseaseVolgaGermanfamiliesincreasesthesecretionofamyloidbetaproteinendingatthe42ndor43rdresidue.ProceedingsoftheNationalAcademyofScience,94,2025-2030.Touchon,J.,Bergman,H.,Bullock,R.,Rapatz,G.,Nagel,J.,Lane,R.006.ResponsetorivastigmineordonepezilinpatientswithAlzheimer'sdiseaseandsymptomssuggestiveofconcomitantLewybodypathology.CurrentMedicalResearchandOpinion,22,49-59.306

PAGE 317

Touzani,K.,Puthanveettil,S.V.,andKandel,E.R.2007.Consolidationoflearn-ingstrategiesduringspatialworkingmemorytaskrequiresproteinsynthesisintheprefrontalcortex.ProceedingsoftheNationalAcademyofSciencesUSA,104,5632-5637.Tranel,D.,Damasio,A.R.,Damasio,H.,andBrandt,J.P.94.Sensorimotorskilllearninginamnesia:Additionalevidencefortheneuralbasisofnondeclar-ativememory.LearningandMemory,1,165-179.Tulving,E.72.Episodicandsemanticmemory.Reprintedin:Tulving,E.,andDonaldson,W.Eds.OrganizationandMemory.NewYork:AcademicPress.pp.381-403.Tulving,E.985.Howmanymemorysystemsarethere?AmericanPsychologist,40,385-398.Ubeyli,E.D.2008.Multiclasssupportvectormachinesfordiagnosisoferythemato-squamousdiseases.ExpertSystemswithApplications,35,1733-1740.Underwood,B.J.,Boruch,R.F.,andMalmi,R.A.1978.Compositionofepisodicmemory.JournalofExperimentalPsychology:General,107,393-419.Valenti,P.,Cazamajou,E.,Scarpettini,M.,Aizemberg,A.,Silva,W.,andKochen,S.2006.Automaticdetectionofinterictalspikesusingdataminingmodels.JournalofNeuroscienceMethods,150,105-110.VanBroeckhovenC,BackhovensH,CrutsM,DeWinterG,BruylandM,CrasP,andMartinJ.J.992.Mappingofagenepredisposingtoearly-onsetAlzheimer'sdiseasetochromosome14q24.3.NatureGenetics,2,335-339.VanDam,D.,DHooge,R.,Staufenbiel,M.,VanGinneken,C.,VanMeir,F.,andDeDeyn,P.003.Age-dependentcognitivedeclineintheAPP23modelprecedesamyloiddeposition.EuropeanJournalofNeuroscience,17,388-396.VanDerStaay,F.J.,andSteckler,T.01.Behavioralphenotypingofmousemutants.BehavioralBrainResearch,125,3-12.VanGelder,B.M.,Buijsse,B.,Tijhuis,M.,Kalmijn,S.,Giampaoli,S.,Nissinen,A.,andKromhout,D.07.CoeeconsumptionisinverselyassociatedwithcognitivedeclineinelderlyEuropeanmen:TheFINEStudy.EuropeanJournalofClinicalNutrition,61,226-232.Vapnik,V.1998.StatisticalLearningTheory.NewYork:Wiley.Vapnik,V.,andChervonenkis,A.964.Anoteononeclassofperceptrons.AutomationandRemoteControl,25,821-837.307

PAGE 318

Vapnik,V.,andLerner,A.1963.Patternrecognitionusinggeneralizedportraitmethod.AutomationandRemoteControl,24,774-780.Vassar,R.,Bennett,B.D.,Babu-Khan,S.,Kahn,S.,Mendiaz,E.A.,Denis,P.,Teplow,D.B.,Ross,S.,Amarente,P.,Loelo,R.,Luo,Y.,Fisher,S.,Fuller,J.,Edenson,S.,Lile,J.,Jarosinski,M.A.,Biere,A.L.,Curran,E.,Burgess,T.,Louis,J.C.,Collins,F.,Treanor,J.,Rogers,G.,andCitron,M.1999.Beta-secretasecleavageofAlzheimer'samyloidprecursorproteinbythetrans-membraneasparticproteaseBACE.Science,286,735-741.Vaughan,W.,Jr.88.Formationofequivalencesetsinpigeons.JournalofExperimentalPsychology:AnimalBehaviorProcesses,14,36-42.Vekovischeva,O.Yu.,Verbitskaya,E.V.,Aitta-aho,T.,Sandnabba,K.,andKorpi,E.R.2007.Multimetricstatisticalanalysisofbehaviorinmiceselectedforhighandlowlevelsofisolation-inducedmaleaggression.BehavioralProcesses.75,23-32.Velicer,W.F.,andFava,J.L.98.Eectsofvariableandsubjectsamplingonfactorpatternrecovery.PsychologicalMethods,3,231-251.Velicer,W.F.,andJackson,D.N.90.Componentanalysisversuscommonfactoranalysis:Someissuesinselectinganappropriateprocedure.MultivariateBehavioralResearch,25,1-28.Venneri,A.,Forbes-Mckay,K.E.,andShanks,M.F.2005.ImpoverishmentofspontaneouslanguageandthepredictionofAlzheimer'sdisease.Brain,128,E27.Vercelli,D.04.Genetics,epigenetics,andtheenvironment:Switching,buer-ing,releasing.JournalofAllergyandClinicalImmunology,113,286-381.Verghese,J.,Lipton,R.,Katz,M.,Hall,C.,Derby,C.,Kuslansky,G.,Ambrose,A.,Sliwinski,M.,andBuschke,H.2003.Leisureactivitiesandtheriskofdementiaintheelderly.NewEnglandJournalofMedicine,348,2508-2516.Veurink,G.,Fuller,S.,Atwood,C.,andMartins,R.03.Genetics,lifestyle,andtherolesofamyloidbetaandoxidativestressinAlzheimer'sdisease.AnnalsofHumanBiology,30,639-667.Viera,A.J.,andGarrett,J.M.2005.Understandinginterobserveragreement:Thekappastatistic.FamilyMedicine,37,360-363.308

PAGE 319

Vinores,S.,Xiao,W.,Zimmerman,R.,Whitcup,S.andWawrousek,E.003.UpregulationofvascularendothelialgrowthfactorVEGFintheretinasoftransgenicmiceoverexpressinginterleukin-1betaIL-1betainthelensandmiceundergoingretinaldegeneration.HistologyandHistopathology,18,797-810.Waddington,C.H.53.Epigeneticsandevolution.SymposiumoftheSocietyforExperimentalBiology,7,186-199.Wahlsten,D.,Cooper,S.F.,andCrabbe,J.C.005.Dierentrankingsofin-bredmousestrainsontheMorrismazeandarened4-armwaterescapetask.BehavioralBrainResearch,165,36-51.Walczak,S.,andCerpa,N.99.Heuristicprinciplesforthedesignofarticialneuralnetworks.InformationandSoftwareTechnology,41,107-117.Walker,P.R.,Smith,B.,Liu,Q.Y.,Famili,A.F.,Valdes,J.J.,Liu,Z.,andLach,B.004.DataminingofgeneexpressionchangesinAlzheimerbrain.ArticialIntelligenceinMedicine,31,137-154.Wall,P.M.,andMessier,C.000.Ethologicalconrmatoryfactoranalysisofanxiety-likebehaviorinthemurineelevatedplus-maze.BehavioralBrainRe-search,114,199-212.Walsh,D.M.,andSelkoe,D.J.004.Oligomersonthebrain:Theemergingroleofsolubleproteinaggregatesinneurodegeneration.ProteinandPeptideLetters,11,213-228.Wang,J.,Ikonen,S.,Gurevicius,K.,vanGroen,T.,andTanila,H.02.Alter-ationofcorticalEEGinmicecarryingmutatedhumanAPPtransgene.BrainResearch,943,181-190.Watanabe,S.,Sakamoto,J.,andWakita,M.1995.Pigeons'discriminationofpaintingsbyMonetandPicasso.JournaloftheExperimentalAnalysisofBe-havior,63,165-174.Webster,S.,Lue,L.F.,Brachova,L.,Tenner,A.J.,McGeer,P.L.,Terai,K.,Walker,D.G.,Bradt,B.,Cooper,N.R.,andRogers,J.1997.Molecu-larandcellularcharacterizationofthemembraneattackcomplex,C5b-9,inAlzheimer'sdisease.NeurobiologyofAging,18,415-421.Weggen,S.,Eriksen,J.L.,Das,P.,Sagi,S.A.,Wang,R.,Pietrzik,C.U.,Findlay,K.A.,Smith,T.E.,Murphy,M.P.,Bulter,T.,Kang,D.E.,Marquez-Sterling,N.,Golde,T.E.,andKoo,E.H.2001.AsubsetofNSAIDsloweramyloido-genicAbeta42independentlyofcyclooxygenaseactivity.Nature,414,212216.309

PAGE 320

Weiner,M.F.,Hynan,L.S.,Bret,M.E.,andWhite,C.005.EarlybehavioralsymptomsandcourseofAlzheimer'sdisease.ActaPsychiatricaScandinavica,111,367-371.Weinhold,B.2006.Epigenetics:Thescienceofchange.EnvironmentalHealthPerspectives,114,A160-A167.Welsh,K.,Butters,N.,Hughes,J.,Mohs,R.,andHeyman,A.991.DetectionofabnormalmemorydeclineincasesofAlzheimer'sdiseaseusingCERADneurologicalmeasures.ArchivesofNeurology,48,278-281.Wesnes,K.,Simpson,P.M.,andKidd,A.G.988.Aninvestigationoftherangeofcognitiveimpairmentsinducedbyscopolamine0.6mg.HumanPsychophar-macology,3,27-41.Westbury,C.,Buchanan,L.,Sanderson,M.,Rhemtulla,M.,andPhillips,L.003.Usinggeneticprogrammingtodiscovernonlinearvariableinteractions.Behav-iorResearchMethods,Instruments,andComputers,35,202-216.Westerman,M.,Cooper-Blacketer,D.,Mariash,A.,Kotilinek,L.,Kawara-bayashi,T.,Younkin,L.,Carlson,G.,Younkin,S.,andAshe,K.002.Therela-tionshipbetweenAandmemoryintheTg2576mousemodelofAlzheimer'sdisease.JournalofNeuroscience,22,1858-1567.Whelihan,W.M.,Thompson,J.A.,Piatt,A.L.,Caron,M.D.,andChung,T.997.Therelationofneuropsychologicalmeasurestolevelsofcognitivefunc-tioninginelderlyindividuals:Adiscriminantanalysisapproach.AppliedNeu-ropsychology,4,160-164.Whishaw,I.Q.,andTomie,J.A.96.Ofmiceandmazes:Similaritiesbetweenmiceandratsondrylandbutnotwatermazes.PhysiologyandBehavior,60,1191-1197.Wilks,S.S.932.Certaingeneralizationsintheanalysisofvariance.Biometrika,24,471-494.Willingham,D.B.,andKoroshetz,W.J.93.EvidencefordissociablemotorskillsinHuntington'sdiseasepatients.Psychobiology,21,173-182.Wilson,R.S.,Bienias,J.L.,Berry-Kravis,E.,Evans,D.A.,andBennett,D.A.002.TheapolipoproteinE2alleleanddeclineinepisodicmemory.JournalofNeurology,NeurosurgeryandPsychiatry,73,672-677.Wilson,R.S.,Li,Y.,Aggarwal,N.T.,Barnes,L.L.,McCann,J.J.,Gilley,D.W.,andEvans,D.A.04.EducationandthecourseofcognitivedeclineinAlzheimer'sdisease.Neurology,63,1198-1202.310

PAGE 321

Wilson,R.S.,Scherr,P.A.,Schneider,J.A.,Tang,Y.,andBennett,D.A.007.TherelationofcognitiveactivitytoriskofdevelopingAlzheimerdisease.Neu-rology,69,1911-1920.Winblad,B.,Engedal,K.,Soininen,H.,verhey,F.,Waldemar,G.,Wimo,A.,Wet-terholm,A.,Zhang,R.,Haglund,A.,Subbiah,P.andtheDonepezilNordicStudyGroup.2001.A1-year,randomized,placebo-controlledstudyofdonepezilinpatientswithmildtomoderateAD.Neurology,57,489-495.Winocur,G.984.Theeectsofretroactiveandproactiveinterferenceonlearningandmemoryinoldandyoungrats.DevelopmentalPsychobiology,17,537-545.Wisniewski,T.,Castano,E.,Golabek,A.,Vogel,T.andFrangione,B.94.Ac-celerationofAlzheimer'sbrilformationbyapolipoproteinEinvitro.Ameri-canJournalofPathology,145,1030-1035.Witten,I.H.,andFrank,E.000.DataMining:PracticalMachineLearningToolsandTechniqueswithJavaImplementations,2ndEd.NewYork:MorganKaufmannPublishers.Wolfer,D.P.,Stagljar-Bozicevic,M.,Errington,M.L.,andLipp,H.-P.998.Spatialmemoryandlearningintransgenicmice:Factorartifact?NewsinPhysiologicalSciences,13,118-123.Wolpert,D.H.,andMacready,W.G.997.Nofreelunchtheoremsforoptimiza-tion.IEEETransactionsonEvolutionaryComputation,1,67-82.Wong,P.C.,Cai,H.,Borchelt,D.R.,andPrice,D.L.02.Geneticallyengi-neeredmousemodelsofneurodegenerativediseases.NatureNeuroscience,5,633-639.Woodru-Pak,D.S.,Vogel,R.W.,andWenk,G.L.01.Galantamine:Ef-fectonnicotinicreceptorbinding,acetylcholinesteraseinhibition,andlearning.ProceedingsoftheNationalAcademyofSciencesUSA,98,2089-2094.Wright,J.W.,Alt,J.A.,Turner,G.D.,andKrueger,J.M.004.Dierencesinspatiallearningcomparingtransgenicp75knockout,NewZealandBlack,C57BL/6,andSwissWebstermice.BehavioralBrainResearch,153,453-458.Wu,J.,Basha,M.R.,andZawia,N.H.008a.Theenvironment,epigeneticsandamyloidogenesis.JournalofMolecularNeuroscience,34,1-7.Wu,J.,Basha,M.R.,Brock,B.,Cox,D.P.,Cardozo-Pelaez,F.,McPherson,C.A.,Harry,J.,Rice,D.C.,Maloney,B.,Chen,D.,Lahiri,D.K.,andZawia,N.H.008b.Alzheimer'sdiseaseAD-likepathologyinagedmonkeysafterinfan-tileexposuretoenvironmentalmetalleadPb:EvidenceforadevelopmentaloriginandenvironmentallinkforAD.JournalofNeuroscience,28,3-9.311

PAGE 322

Wyss-Coray,T.006.InammationinAlzheimerdisease:Drivingforce,by-standerorbenecialresponse?NatureMedicine,12,1005-1015.Xiao,Q.X.,Wang,R.,andXi,C.T.00.HuperzineAandtacrineattenu-atebeta-amyloidpeptide-inducedoxidativeinjury.JournalofNeuroscienceResearch,61,564-569.Xu,Y.,Jack,C.R.,OBrien,P.C.,Kokmen,E.,Smith,G.E.,Ivink,R.J.,Boeve,B.F.,Tangalos,R.G.,andPetersen,R.C.2000.UsefulnessofMRImeasuresofentorhinalcortexversushippocampusinAlzheimer'sdisease.Neurology,54,1760-1767.Yamaguchi,H.,Hirai,S.,Morimatsu,M.,Shoji,M.,andHarigaya,Y.988.DiusetypeofsenileplaquesinthebrainsofAlzheimer-typedementia.ActaNeuropathologica,77,113-119.Yao,J.,Petanceska,S.S.,Montine,T.J.,Holtzman,D.M.,Schmidt,S.D.,Parker,C.A.,Callhan,M.J.,Lipinski,W.J.,Bisgaier,C.L.,Turner,B.A.,Nixon,R.A.,Martins,R.N.,Ouimet,C.,Smith,J.D.,Davies,P.,Laska,E.,Ehrlich,M.E.,Walker,L.C.,Mathews,P.M.,andGandy,S.004.Aging,gen-der,andApoEisotypemodulatemetabolismofAlzheimer'sApeptidesandF-isoprostanesintheabsenceofdetectableamyloiddeposits.JournalofNeu-rochemistry,90,1011-1018.Yoshikai,S.,Sasaki,H.,Doh-ura,K.,Furuya,H.,andSakaki,Y.990,Genomicorganizationofthehumanamyloidbeta-proteinprecursorgene.Gene,87,257-263.Zandi,P.P.,Anthony,J.C.,Hayden,K.M.,Mehta,K.,Mayer,L.,Breitner,J.C.,andCacheCountyStudyInvestigators.02.ReducedincidenceofADwithNSAIDbutnotH2receptorantagonists:TheCacheCountyStudy.Neurology,59,880-886.Zangara,A.2003.ThepsychopharmacologyofhuperzineA:AnalkaloidwithcognitiveenhancingandneuroprotectivepropertiesofinterestinthetreatmentofAlzheimer'sdisease.PharmacologyBiochemistryandBehavior,75,675-686.Zhang,G.P.00.Neuralnetworksforclassication:Asurvey.IEEETransac-tionsonSystems,Man,andCybernetics,PartC:ApplicationsandReviews,30,451-462.Zhuo,J.-M.,Prescott,S.L.,Murray,M.E.,Zhang,H.-Y.,Baxter,M.G.,andNicolle,M.M.07.EarlydiscriminationreversallearningimpairmentandpreservedspatiallearninginalongitudinalstudyofTg2576APPswmice.Neu-robiologyofAging,28,1248-1257.312

PAGE 323

Zillmer,E.A.,Fowler,P.C.,Gutnick,H.N.,andBecker,E.90.Comparisonoftwocognitivebedsidescreeninginstrumentsinnursinghomeresidents:Afactoranalyticstudy.JournalofGerontology,45,P69-P74.Zimmer,C.08,Oct.Thesearchforintelligence.ScienticAmerican,2994,68-75.Zimmermann,M.,Colciaghi,F.,Cattabeni,F.,andDiLuca,M.02.Ginkgobilobaextract:FrommolecularmechanismstothetreatmentofAlzheimer'sdisease.CellularandMolecularBiology,48,613-623.Zubenko,G.S.,Hughes,H.B.,III,andStier,J.S.001.D10S1423identiesasusceptibilitylocusforAlzheimer'sdiseaseinaprospective,longitudinal,double-blindstudyofasymptomaticindividuals.MolecularPsychiatry,6,413-419.Zweig,M.H.,andCampbell,G.1993.Receiver-operatingcharacteristicROCplots:Afundamentalevaluationtoolinclinicalmedicine.ClinicalChemistry,39,561-577.Zwick,W.R.,andVelicer,W.F.986.Comparisonofverulesfordeterminingthenumberofcomponentstoretain.PsychologicalBulletin,99,432-442.313

PAGE 324

ABOUTTHEAUTHORRalphE.LeightyattendedtheUniversityofFloridaGainesvilleasanunder-graduate,wherehecompletedBachelor'sdegreesinChemistry1985andPsychol-ogy988,followedbypostbaccalaureatestudyinPharmacyandNeurobiologicalSciences.In1991,herelocatedtoTampa,FloridatopursuegraduatestudiesinExperimentalPsychologyattheUniversityofSouthFlorida.HetransferredtotheCollegeofEngineeringin1995tostudyarticialintelligenceandrobotics,andre-ceivedaMaster'sinComputerScience997.In2001,heresumedhisgraduatestudies,earningaMaster'sinBiology03withathesisonbehavioralanalysisofstraindierencesandAlzheimer-liketransgeniceectsinmice.Hiscurrentresearchinterestsincludebehavioralneurobiology,computationalneuroscience,andnonlinearanalysis.Inadditiontohisacademicactivities,heisanactor/comedian,musicianpiano,synthesizers,andchessenthusiast.HecurrentlyresidesinTampawithhiswife,Debby,andtheirtwocats,JoandMeg.