REVIEWARTICLE
Aclassificationoftasksforthesystematicstudyofimmuneresponseusingfunctionalgenomicsdata
C.HEDELER1,2*,N.W.PATON1,J.M.BEHNKE3,J.E.BRADLEY3,M.G.HAMSHERE3andK.J.ELSE212SchoolofComputerScience,TheUniversityofManchester,OxfordRoad,ManchesterM139PL,UK
FacultyofLifeSciences,MichaelSmithBuilding,TheUniversityofManchester,OxfordRoad,ManchesterM139PT,UK3SchoolofBiology,NottinghamUniversity,NottinghamNG72RD,UK
(Received30November2004;revised25Marchand23June2005;accepted30June2005;firstpublishedonline21September2005)
SUMMARY
Afullunderstandingoftheimmunesystemanditsresponsestoinfectionbydifferentpathogensisimportantforthedevelopmentofanti-parasiticvaccines.Agrowingnumberoflarge-scaleexperimentaltechniques,suchasmicroarrays,arebeingusedtogainabetterunderstandingoftheimmunesystem.Toanalysethedatageneratedbytheseexperiments,methodssuchasclusteringarewidelyused.However,individualapplicationsofthesemethodstendtoanalysetheexperimentaldatawithouttakingpubliclyavailablebiologicalandimmunologicalknowledgeintoaccountsystematicallyandinanunbiasedmanner.Tomakebestuseoftheexperimentalinvestment,tobenefitfromexistingevidence,andtosupportthefindingsintheexperimentaldata,availablebiologicalinformationshouldbeincludedintheanalysisinasystematicmanner.Inthisreviewwepresentaclassificationoftasksthatshowshowexperimentaldataproducedbystudiesoftheimmunesystemcanbeplacedinabroaderbiologicalcontext.Takingintoaccountavailableevidence,theclassificationcanbeusedtoidentifydifferentwaysofanalysingtheexperimentaldatasystematically.Wehaveusedtheclassificationtoidentifyalternativewaysofanalysingmicroarraydata,andillustrateitsapplicationusingstudiesofimmuneresponsesinmicetoinfectionwiththeintestinalnematodeparasitesTrichurismurisandHeligmosomoidespolygyrus.Keywords:classification,systematicimmunologicalbioinformatics,intestinalnematode.
INTRODUCTION
Thestudyofimmuneresponsestoinfectionbypathogensprovidesusefulinsightsforthedevelop-mentofanti-parasiticvaccines.Theimmunesystemiscapableofmountingdifferenttypesofresponsesthatconsistofdifferentphasesandmechanisms,suchasimmediateanddelayedresponses.Thismakesithardtounderstanditcompletelyinitscomplexity.Thetypeofresponsemountedbytheimmunesystemcandependonseveraldifferentfactorsoroncombinationsofthosefactors.Examplesofthesefactorsarethegeneticbackgroundofthehost(ElseandWakelin,1988),thetypeofpathogenandthestrain/isolateofpathogen(Bellaby,RobinsonandWakelin,1996),orthedoselevelwithwhichthehosthasbeeninfected(Bretscheretal.1992;Bancroft,ElseandGrencis,1994),tomentionbutafew.
Togainabetterunderstandingoftheimmunesystem,themouseMusmusculusiswidelyusedasa
*Correspondingauthor:SchoolofComputerScience,TheUniversityofManchester,OxfordRoad,ManchesterM139PL,UK.Tel:+44(0)1612757821.Fax:+44(0)1612756236.E-mail:chedeler@cs.man.ac.uk
modelorganism.Withtheavailabilityofdifferentstrainsandgene-targetedknock-outmice,itcanbeusedtostudyindetaildifferentaspectsorstagesoftheimmuneresponsetoinfection(Mak,PenningerandOhashi,2001).
Insuchcontext,agrowingnumberofanalyticaltechniquesareapplied.Thesetechniquesrangefromthehypothesis-drivensmallscale,suchasWesternimmunoblots,tothecollection-drivenlargescale,suchasmicroarrays,oneoftheemergingtechniquesinthepost-genomicera.Large-scaletechniquesarealsocalledhigh-throughputtechniques.Theycanbeusedtotesthypothesesand,duetotheirscale,canalsobeusedtogenerateorrefinehypotheses.Thesecanthenbetestedmorethoroughlybysmall-scaletechniques.Thecomplementaryuseofbothtypesofanalysistechniquesformsaniterative‘cycleofknowledge’(KellandOliver,2003).
Tobenefitfromhigh-throughputexperiments,thevastamountsofdataproducedbythesetechniquesneedtobeanalysed.Thiscanbedonebyfilteringthedatatoeliminatelow-qualitymeasurements,normalization(e.g.forareviewofanalysismethodsfortranscriptomedataseeQuackenbush(2002)),andidentificationofthegenesorproteinsofinterest.
Parasitology(2006),132,157–167.f2005CambridgeUniversityPressdoi:10.1017/S0031182005008796PrintedintheUnitedKingdom
C.Hedelerandothers158
Table1.Analysistechniquesusedintheliterature
(Analysistechniquesusedintheliteraturepublishedongeneexpressionstudiesusedtostudytheimmuneresponsetoinfectionbypathogens.)Reference
Langetal.(2003)Crokeretal.(2003)Muelleretal.(2003)Edwardsetal.(2003)Jietal.(2003)Bystro¨metal.(2004)
Domachowskeetal.(2002)Hoffmannetal.(2001)Bladeretal.(2001)
StatisticaltestXX
X
XX
XXXX
Ranking,thresholdXX
X
XXX
XX
X
Clustering
Correlating
Toidentifythegenesofinterestinatranscriptomeexperimentseveraldifferentapproachescanbeused,suchasidentifyingdifferentiallyexpressedgenesbasedontheirfold-changeorbyusingstatisticaltests(Pan,2002).Furthermore,supervisedorun-supervisedclusteringtechniquescanbeappliedtoclustergeneswithsimilarexpressionpatterns(Sherlock,2000).Theexperimentaldatacanalsobeplacedintobiologicalcontextbycorrelatingthedatatootherinformation,suchasfunctionalannotation,chromosomallocationorinformationaboutpath-ways.Boththefold-changeapproachandstatisticaltestshavebeenmainlyusedinstudiesoftheimmuneresponse(Table1).
Usingstatisticaltestsorthefold-changeapproachtoidentifydifferentiallyexpressedgenes,simplyre-ducesthenumberofgenesthathavetobeconsideredforfurtheranalysis.However,byexcludinggenesfromfurtheranalyses,thisapproachmightevenignoreinformationthatcanprovetobevaluablewhenplacedintobiologicalcontext.Moreover,microarrayexperimentsareoftenusedasastartingpointforfurtherexperiments,forinstance,useofknock-outmice,studyofdifferenttime-points,ortostatehypothesestobetested,thenusinghypothesis-drivenanalysistechniques.Forexample,Blader,MangerandBoothroyd(2001)identifiedgenessofarnotknowntobeinvolvedintheimmuneresponsetoinfectionwithToxoplasmagondiiandconfirmedtheresultsusingNorthernBlots.Bystro¨metal.(2004)identifiedgenesexpectedtobeinvolvedinimmuneresponsetoinfectionwithSchistosomamansoni,butforwhichnochangeinexpressionlevelswasobserved.Thisledtonewspeculationsthatrequireexperimentalassessment.Thefindingswerecon-firmedusingRT-PCR.
Toexploitthefullpotentialofsuchexperiments,makeunbiasedobservations,andgainmoreinsightsintotheimmunesystemusingaholisticapproachratherthanstudyingeachcomponentorparameterseparately(Ricciardi-CastagnoliandGranucci,2002),high-throughputdataneedtobeanalysedandcorrelatedsystematicallywithavailablebiological
knowledge(NoordewierandWarren,2001).Examplesofthisknowledgearechromosomallocation,SingleNucleotidePolymorphisms(SNPs),functionalannotationofgenes,pathwaysrelevanttothegenesinvolved,andresultsofotherhigh-throughputstudies.
Toaddressthisneed,wehavedevelopedaclassi-ficationofanalyticaltasksinimmunologicalbio-informaticsinthecontextofimmuneresponsetoinfection.Theclassificationprovidesdifferentwaystoanalyseexperimentaldatainasystematicmannerandtoplaceitinabiologicalcontext.Inthisreview,weintroducetheclassificationandillustrateitwithreferencetoastudyoftheimmuneresponseinthemousetoinfectionwiththeintestinalparasiteTrichurismuris.Thenweshowpossiblewaysofdeploymentoftheclassification,forinstance,toidentifydifferentapproachesofanalysingexper-imentaldata.
THECLASSIFICATION
Toidentifytheanalyticaltasksofrelevancetoimmunologyinthefunctionalgenomicsera,acombinationofbottom-upandtop-downapproacheshasbeenused(seealsoFig.1).
Thebottom-upapproachcanbeseenasdata-driven.Startingwiththeidentificationoftherelevantdata,severalsimpleanalysistasksthatcanbecarriedoutonthesedatasourceshavebeenident-ified.Thesetaskscanbecomposedfurthertoformmorecomplexandcontext-richanalysesandtocombineinformationfromseveraldatasources.Theseanalysesaresimpleinferences,targetedatextractingspecificlessonsfromoneorasmallnumberofexperiments.
Incontrasttothesimpleanalysesandtheircom-positions,themoregeneralandcomplexanalysistasksaredrivenbyimmunologicalknowledgeusingthetop-downapproach.Thesetasksarecomplexinferencestargetedatlearningagenerallesson.Thehigherlevelanalyseshavebeenclassifiedbyassoci-atingthemingroupswithregardtotheircontents.
Aclassificationoftaskstostudyimmuneresponse
SusceptibilityInfected hostInfecting agentImmunology159
Top-downImmunology-drivenWhy is one mouse susceptibleand another one not?What did the immune system of thesusceptible mouse do inappropriately?Did the pathogen trigger the immuneresponse in a direction inappropriatefor the host to ensure its own survival?Which proteins, secreted by thepathogen and similar to proteins inthe host, are essential to ensure thepathogens survival?GeneralquestionsWhich of the strain differencesfound between susceptible andresistant mice are significantfor susceptibility/resistance?Which of the differently activatedpathways in susceptible/resistantmice are significant?BiologicallessonsquestionsWhich strain differences can befound between susceptible andresistant mice?Which genes in a QTL regionand/or with SNPs/differenthaplotypes are differentlyexpressed in susceptible/resistant strains?Which pathways are differentlyactivated in susceptible and resistantmice?Looking at differently expressedgenes in susceptible and resistantmice: in which pathways are theyinvolved?What are the expression levels ofgenes involved in a particular pathway?Which proteins does the pathogensecrete that are similar to proteinsin the host?ComplexquestionsAre there protein-protein interactionsin host and pathogen that are similarto each other? Are they involved insimilar pathways in host and pathogen?Compositionof simplequestionsBottom-upData-drivenWhich geneshave SNPs/differenthaplotype/are ina QTL region?Which data setsfrom susceptible/resistant strainsare there?Which genes aredifferentlyregulated in them?Which proteinshave beenidentified on aparticular 2Dgel map?In which pathwaysis a set of genes/proteins involved?Which genes areinvolved in aparticular pathway?Which protein-protein interactionscan be found in thehost?Which ones in thepathogen?In which pathwaysis a set of genes/proteins involved?Which genes areinvolved in aparticular pathway?SimplequestionsGenomeTranscriptomeProteomeMetabolomeInteractomeControlDatasourcesFig.1.Overviewclassification.Schematicoverviewoftheapproachusedtoclassifythetasks,includingsomeexamplesoftheresultingclassification.Thebottomrowcontainsdifferentkindsofavailabledatathatdrivethesimplequestions,whereastherowsatthetopspecifydifferentaspectsofimmunologicalstudiesthatdrivethemorecomplexquestions.
Thesegroupsare:thereactionofthehosttoaninfection,theinfectingagent,andthereasonsfortheoverallsusceptibilityofthehost.
Theanalysistasks,datasourcesofrelevancetoimmunology,andthetaskclassificationareexplainedinmoredetailbelow.Classificationofthedata
Availableandrelevantdata,includingexperimentaldata,havebeenclassifiedaccordingtotheircontents,resultinginthefollowingcategories:Genome,Transcriptome,Proteome,Metabolome,Inter-actome,andControl.AnoverviewoftheresultingcategoriesandasubsetofthedatasourcesusedareshowninTable2.
Classificationofimmunology
Toidentifythemorecomplexquestions,drivenbyimmunologicalknowledge,aclassificationofimmunologyisrequired.Thefollowingdifferentaspectsofimmunologicalstudyandinteresthavebeenidentified:studyofthehostpost-infection,ofthepathogenpost-infectionandofthesusceptibilityofanindividualtoinfectionorre-infectionwithaparticularpathogen.Theinfectedhostmountsanimmuneresponsethatcanconsistofdifferentstages.Theseincludethedetectionofinfectionandtheimmediateanddelayedresponsetoinfection.Theseresponsesresultineitherthedestructionofthepathogen,neutralizationofthethreatandprovisionofimmunity,ortheenteringofanalteredstatetopreventhost-damagingpathology.Thelattermayoccurinthecaseofchronicinfections.
Theinfectingagentinitiallyinvadeshosttissue.Thisisfollowedbyanevasionoftheimmuneresponseand,ontreatmentofthehostwithdrugs,byaresponsetothesedrugs.
Asindicatedabove,thetypeofimmuneresponsemountedbytheimmunesystemdependsonseveralfactors.Thesecancausedifferencesinex-pressionlevelsandleadtodifferentactivationsofpathwaysresultingindifferenttypesofresponse(Fig.2).
However,differencesingeneexpressionlevelscannotonlybecausedbydifferentpathogensordifferentstrainsofthehost.Theycanalsobecausedbychangesintheexperimentalconditions,forinstance,thetissuetype,celltypeorthestageoftheimmuneresponse(time-pointpost-infection)examined.Therefore,itisnecessarytotakeallthesedifferentfactorsanddependenciesintoaccount
C.Hedelerandothers160
Table2.Datacategories
(Representativedatacategoriesandasubsetofdatasourcesthatareofrelevancetoimmunology.Amoredetailedoverviewcanbefoundinthesupplementarydatafile1.)CategoryofdataGenome
DatainthiscategorySequenceLocationStrain
Functionalannotation
TranscriptomeProteomeMetabolomeInteractomeControl
SpeciescomparisonMicroarraydescriptionExperimentalconditionResult
ProteomicsexperimentaldataMetabolicpathways
Protein-proteininteractionCellular,molecularandregulatorypathways
Datasources
ENSEMBL(Hubbardetal.2002)ENSEMBL,MGD(Blakeetal.2003)
ENSEMBL,dbSNP(Wheeleretal.2003),MGD(Eppigetal.2002)
GO(TheGeneOntologyConsortium,2000),
MGD,ENSEMBL,InterPro(Mulderetal.2003)ENSEMBL
Locallyproducedexperiments,SMD
(Gollubetal.2003),GEO(Wheeleretal.2003)SWISS-2DPAGE(http://ca.expasy.org/ch2d/)KEGG(Kanehisaetal.2002)
BIND(Bader,BetelandHogue,2003),DIP(Xenariosetal.2002)
BioCarta(http://www.biocarta.com),KEGG
Strain differences
in the hostDifferentpathogensDifferent life cyclestages of pathogenStrain/isolate differences
in the pathogenDifferent dose
levelsFirst or secondtime of infection
Differences in gene expression levels(different set of genes up-/down-regulated;same or similar set of genes differently regulated)
Different pathways
Whole pathway different/only partly different/different set of pathways
Different (type of)immune response
Fig.2.Causefordifferentimmuneresponses.Schematicpresentationoffactorsthatcancausedifferencesingenesexpression,whichinturncanleadtodifferencesinactivationofpathwaysandcancausedifferenttypesofimmuneresponses.
whileanalysingexperimentaldataordevelopingaclassificationoftasks.Classificationofthetasks
ThissectionexaminesthetasksintroducedinFig.1inmoredetail,inparticularthetop-downapproach.Thiswillillustratehowgeneralquestionscanbedecomposedintosimplerrequeststhatcanthenbeansweredbyusingspecificexperimentaldatasets.Question1.Whyisonemousesusceptibleandanotheronenot?Forexample,AKRmicearesusceptibletoinfectionbyTrichurismuriswhileBALB/cmiceareresistant(DeschoolmeesterandElse,2002).Thismightbecausedbystraindifferences,suchaspoly-morphisms.However,therecouldbeseveralSNPsbetweenasusceptibleandresistantstrainandprob-ablynotallofthemareingenesthatareinvolvedinhost-protectiveimmuneresponses.Thisleadstothenexttierofquestions.Whichstraindifferencescanbefoundbetweensusceptibleandresistantmice?Whichofthestraindifferencesfoundbetweensusceptibleandresistantmicearesignificantforsusceptibility/resistance?
Toanswerthesequestions,differencesinthesestrainshavetobeidentifiedbyanalysinggenomedatacontaininginformationaboutpolymorphisms.
Aclassificationoftaskstostudyimmuneresponse161
Transcriptomefind genes differentlyexpressed in susceptibleand resistant strainsgenes differentlyexpressed in susceptibleand resistant strainsformintersectiongenes that are differentlyexpressed and containSNPs in susceptibleand resistant strainsGenomefind genes with SNPsbetween susceptible andresistant strainsgenes with SNP betweensusceptible and resistantstrainsFig.3.Dataflowdiagram–susceptibility.Dataflowdiagramforretrievinggenesthataredifferentlyexpressedin
susceptibleandresistantmicethatcontainSNPs.Thefollowingnotationisused:open-endedrectanglesrepresentdatastores,ellipsesrepresentprocessesthatprocesstheincomingdataandproduceanoutput,andarrowsrepresentthedataflow.
Transcriptomefind genes differentlyexpressed in susceptibleand resistant strainsgenes differentlyexpressed in susceptibleand resistant strainsMetabolomefind the pathways inwhich the differentlyexpressed genes areinvolvedControlpathways in whichgenes that are differentlyexpressed in susceptibleand resistant strainsare involvedFig.4.Dataflowdiagram–host.Dataflowdiagramforretrievinggenesthataredifferentlyexpressedinsusceptibleandresistantmice,andforfindingthepathwaysthesegenesareinvolvedin.
However,theresultofthistaskwillbealargenumberofpolymorphicgenes.Toreducethelisttogenesinvolvedinthehost-protectiveimmuneresponse,thegeneswithdifferentexpressionlevelsinsusceptibleandresistantstrainscanbechosen.Theresultingsetofgenesprobablydoesnotprovideenoughinformationtoanswerthemoregeneralquestions.However,analysingdifferentbatchesofappropriateexperimentaldatasetsusingcompo-sitionsofsimpletasksmighteventuallyleadtotheanswer(forasystematicoverviewseeFig.3).Thisapproachcanbeimprovedfurtherbytakingintoaccountpolymorphismsinthestructuralandpromoterregionsofgenes.Polymorphismsintheseregionswillalsoinfluencetheresistanceorsuscep-tibilityofthehost.
Question2.Whatdidtheimmunesystemofthesusceptiblemousedoinappropriately?Itisknown,forinstance,thatmicesusceptibletoinfectionbyT.murismountaninappropriateTh1immuneresponse.However,resistantmicemountaTh2responseandexpelthewormbeforeday35post-infection(DeschoolmeesterandElse,2002).Bothimmuneresponsesconsistofseveralpathways;however,itisnotyetknowwhetherjusttheTh1andTh2signallingpathwaysareimportantorwhetherotherfactorsplayaroletoo.
Therefore,toanswerQuestion2thepathwaysthataredifferentlyactivatedinresistantandsusceptiblemiceneedtobestudied.Thismightbedonebyanalysingseveraltranscriptomedatasets,findingthegenesthataredifferentlyregulated,andidentifyingthepathwaystheyareinvolvedin.ThedataflowdiagramforthistaskisshowninFig.4.Afteridenti-ficationofthepathwaysofinterest,thesignificantpathwaysamongtheseneedtobeidentified,whichmightrequiretheanalysisofmoredatasets.
C.Hedelerandothers
Control162
MetabolomeInteractomefind protein-proteininteractions in the hostproteins involvedin protein-proteininteraction in thehostfind the pathways inwhich the proteinsare involved inpathways in the hostdifferences orsimilarities inprotein-proteininteraction andpathways in hostand pathogenfind differences orsimilarities betweenprotein-proteininteractions andpathways in host andpathogenpathways in the pathogenInteractomefind protein-proteininteractions in thepathogenproteins involvedin protein-proteininteraction in thepathogenfind the pathways inwhich the proteinsare involved inControlMetabolomeFig.5.Dataflowdiagram–pathogen.Dataflowdiagramforidentifyingdifferencesorsimilaritiesinprotein-proteininteractionsandpathwaysinhostandpathogen.
Question3.Didthepathogentriggertheimmuneresponseinadirectionthatisinappropriateforthehost,butthatensuresthesurvivalofthepathogen?Forexample,itisknownthatT.murissecretesaproteinthatissimilartoIFNcinthehost.ThiscancausethehosttomountaTh1response,whichisinappropriateforwormexpulsion(Grencis,2001).Therefore,thequestionmightbeansweredbycomparingthepro-teinssecretedbythehostwiththeonessecretedbythepathogen.Then,thosethataresignificanttoensuringthesurvivalofthepathogencouldbeidentified.Again,thisquestioncanprobablyonlybeansweredbyanalysingalargenumberofappropriatedatasets,butsomeinsightsmightbegainedbyusingtheprocedureshowninFig.5.Onelimitationofthisapproachis,however,thatitwillmisshostparasiteinteractionswhichinvolvecarbohydrates,glyco-lipidsorprocessedproteins.
Thequestionslistedaspartoftheseanalysesformasmallsubsetofquestionsthatcouldbeusedtoanalysethesedatainabroadercontext.Supplementarydatafile2providesamorecompre-hensive,thoughbynomeanscomplete,collectionofquestions.Thequestionsareclassifiedaccordingtotheircomplexityandtypewithrespecttothekindofdataanalysedortheaspectofimmunologystudied.
CASESTUDIES
Thissectiondescribestheexperienceofdeployingtheclassificationfortheanalysisofexperimentaldatageneratedthroughstudiesoftheimmuneresponseinmicetoinfectionwithpathogens.
Casestudy1–Analysisinatop-downmannerAsshownintheprevioussectionandinFig.1,tostudythesusceptibilityofahost,onecanask‘‘Whyisonemousesusceptibleandanotheronenot?’’Forexample,CBAmicearesusceptibletoinfectionwiththegastrointestinalnematodeparasiteHeligmosomoidespolygyrus,whereasSWRmiceareresistant.Asmentionedbefore,thismightbecausedbygeneticdifferencesbetweenthetwostrains.AsshowninFig.1,andfollowingthedataflowdiagramforthisanalysisinFig.3,apossibleapproachtoidentifyingthosegeneticdifferencesistoidentifygeneswithSNPs.Unfortunatelyonlylimitedinfor-mationaboutSNPsiscurrentlyavailableinpubliclyavailabledatabases.
However,quantitativetraitloci(QTL)analysis(RognerandAvner,2003)providesapowerfultechniquefortheidentificationofchromosomalregionsthatcontributetoaparticularphenotypeandmayshowgeneticdifferencesbetweenthetwo
Aclassificationoftaskstostudyimmuneresponsestrains.QTLanalysishasbeenusedtoidentifylociinfluencingtheimmuneresponsetoinfectionwithH.polygyrus(Mengeetal.2003)andprovidestheanswertothefollowingcomplexquestionthatispartoftheanalysisofsusceptibilityingeneral(seeFig.1)‘‘Whichstraindifferencescanbefoundbetweensusceptibleandresistantmice?’’
TheidentifiedQTLregionscanspanseveralcM,andcontainseveralhundredgenes.Thismakesitdifficulttoidentifypotentialcandidategenes.Forexample,theQTLanalysisofSWRandCBAmicetoinfectionwithH.polygyrushasidentifiedQTLregionsonchromosomes1,2,4,8,9,10,11,12,13,17,18,and19(Mengeetal.2003).Inthefollowing,2oftheQTLregionsinwhichMengeandcoworkersidentifiedcandidategenesareusedtoillustratetheapplicationoftheclassification.Oneofthe2QTLregionsidentifiedonchromosome1islocatedbetween15–43cMandcontainsthecandidategenesStat4,CD28andIL1receptors.TheQTLregiononchromosome17,locatedbetween15–45cM,containsthecandidategenesTnfa,mastcellproteases6and7,trefoilfactors1-3andgenesencodingthemajorhistocompatibilitycomplex.
Alloftheseregionscontainlargenumbersofgenes.Someofthemareknowntobeinvolvedinimmuneresponse,includingthecandidategenes,someofthemnotknowntobeinvolved,andsomeofthemevenwithoutaknownfunction.Itislikelythatneitherallofthesegenes,noronlythecandidategenes,aresignificantforthedifferentoutcomesofinfectioninthetwomousestrains.However,withoutanyfurtherinformationitisdifficulttoanswerthefollowingbiologicallessonsquestionofthisanalysis(seeFig.1)‘‘Whichofthestraindifferencesfoundbetweensusceptibleandresistantmicearesignificantforsusceptibility/resistance?’’
Toanswerthisquestion,thegenesintheidentifiedQTLregionsneedtobestudiedfurtherandtheirroleintheimmuneresponseneedstobeanalysed.DependingonthenumberofgenesintheseQTLregions,thismightbetimeconsumingandnotveryefficient.Itmaybeusefultonarrowdownthenumberofgenesthatneedtobecorroboratedwithfurtherexperimentalanalysis,whichcanbedonebycorrelatingtheinformationaboutQTLregionswithtranscriptomedatageneratedtostudythesameinfection.
Therefore,followingthedataflowdiagramfortheanalysisofsusceptibilityingeneralinFig.3,genesthataredifferentlyexpressedinsusceptibleandresistantmiceneedtobeidentified.Athresholdof2.5-foldchangewasusedtoanalysemicroarrayex-perimentscarriedouttostudytheimmuneresponseofmicetoinfectionwithH.polygyrus(Bradley,Behnke,Hamshere,unpublishedobservations).Thisrevealedthatmorethan1000genesaredifferentlyexpressedinguttissueatday35post-infectioninCBAandSWRmice.
163
AnintersectionofdifferentlyexpressedgeneswiththesetofgenesintheQTLregiononchromosome1revealsthatofthecandidategenesonlyStat4showsdifferencesinexpressionlevelsabove2.5-fold.However,lookingattheexpressionlevelsofothergenesinthisQTLregionshowsthatIl18rap,Il18r1,Icos,Stat1,andIl1rl1aredifferentlyexpressedbetweensusceptibleandresistantmice.Thesegenesarenotmentionedascandidategenes(Mengeetal.2003).ThesameanalysiswasalsousedtoanalysetheQTLregiononchromosome17.Thisrevealedthatofthecandidategenes,notonlyH2-Eb1,H2-M3,H2-Ob,H2-DMb1,andTff2,butalsoAif1,Apobec2,PtcraandApomshowdifferentexpressionlevelsinsusceptibleandresistantmice.
Thisanalysisshowsthatonlyafairlysmallnumberofthecandidategenesshowsignificantlydifferentexpressionlevelsbetweensusceptibleandresistantmice.However,italsoshowsthatsomeothergenes,notyetconsideredaspossiblecandidategenes,havesignificantdifferencesinexpressionlevelsbetweenthetwomousestrains.Thisnewknowledgemightleadtorevisionofthelistofcandidategenes.However,withoutcorrelatingtheinformationaboutQTLregionswithtranscriptomedata,thechoiceofcandidategenesisusuallybiasedtowardsgenesthatareknowntobeinvolvedinimmuneresponse.Thislimitsthechancesofidentifyinggenesthatarenotyetknowntobeinvolvedinimmuneresponsebutmightplayaroleintheresponsetoinfectionwithaparticularpathogen.
Inthiscasestudy,wehaveshownthatusingtranscriptomedatacanbroadentheviewbyinclud-inggenesthatarenewinthecontextofimmuneresponse.Placingtranscriptomeandotherexper-imentaldatainabroaderbiologicalcontextbycorrelatingsuchdatawithotherinformationcanhelppreventtheamountofavailableinformationbecomingoverwhelming.Furthermore,suchanapproachcanshowpossibledirectionsforfurtheranalyses.
Casestudy2–Analysisinabottom-upmannerTheclassificationcanalsobeusedbystartingwithaparticularkindorseveralkindsofavailabledata,toidentifywaystoqueryandcombinethesedata.Thisapproachcanbeusedtoanalysethedatainasystematicmannerandtoidentifynovelkindsofanalysesthatmightprovidenewinsights.
Forinstance,analysingmicroarraydatabyiden-tifyinggeneswithsignificantlydifferentexpressionlevelsanswersoneofthesimplequestionsthatcanbeusedtoanalysetranscriptomedata(Fig.1).Thisapproachandathresholdof2.5-foldchangewereusedtoanalysemicroarrayexperimentscarriedouttostudytheimmuneresponseofmicetoT.muris.Theanalysisrevealedthat107genesaredifferentlyexpressedinAKRandBALB/cmiceonday19
C.Hedelerandotherspost-infectioninthegutwith49genesbeingdown-regulatedand58genesbeingup-regulated.Theup-regulatedgenesincludeCasp1,Casp4,Casp8,andCycs.Thesameanalysishasbeenusedfortranscriptomedataofthemesentericlymphnode(MLN).Thisshowsthatatday19post-infection,163genesaredifferentlyexpressed.Ofthese,74genesaredown-regulatedand89genesareup-regulated,includingCasp1andGzmb.However,withoutcorrelatingthisinformationtobiologicalknowledge,thismightnotprovideenoughinfor-mationtounderstandtherelationshipsamongstthesegenes.
Byusingaslightlydifferentapproach(startingfromadifferentpointofview)toanalysethetran-scriptomedataandfocusingongenesinaparticularpathway,onemightbeabletoplacetheobservedchangesinexpressionlevelsinabiologicalcontext.Insteadofstartingwithananalysisofthemicroarraydatabyexcludinggeneswithexpressionlevelsbelowacertainthreshold,thefollowingquestionhasbeenusedtoidentifyallgenesinvolvedinaparticularpathwayofinterest‘‘Whichgenesareinvolvedinaparticularpathwaye.g.caspasecascadeinapoptosis?’’
Thisisoneofthesimplequestionsthatcanbeusedtoanalysemetabolomeandcontroldatasets(seeFig.1).Thisanalysisshowsthatthefollowingareallinvolvedinthecaspasecascadeinapoptosispathway:Adprt1,Apaf1,Arhgdib,Birc2,Birc3,Birc4,Casp1,Casp2,Casp3,Casp4,Casp7,Casp8,Casp9,Cycs,Dffa,Gzmb,Lmna,Lmnb1,Lmnb2,andPrf1.FollowingtheanalysisofmetabolomeandcontroldatashowninFig.1,andcorrelatingthisinformationwithexpressionlevelsofgenes,leadtothefollowingquestion,acompositionofsimplequestions,tobeasked‘‘Whataretheexpressionlevelsofallgenesinvolvedinaparticularpathwaye.g.caspasecascadeinapoptosisinaparticulardataset?’’
Usingthisquestiontoanalysetheexpressionlevelsofgenesinvolvedinthispathwayatday19postinfectioningutandMLNrevealsthefollowing.Casp1,Casp4,Casp8,Cycs,andGzmbareup-regulatedinthegutbuttheexpressionlevelofGzmbisbelowtheappliedthresholdof2.5-foldchange.InMLN,thefollowinggenesareup-regulated:Casp1,Gzmb,Casp4,Arhgdib,Lmnb2,Cycs,Casp8,Birc4,andAdprt1.Thefirsttwoareregulatedabovetheappliedthresholdwhiletheremainderarebelow.
Thisanalysisoftranscriptomedatashowsthatthereisadifferenceinexpressionlevelsingenesinvolvedinthecaspasecascadeintheapoptosispathwayatday19post-infection.However,only4genesinthispathwayhaveanexpressionlevelabove2.5-foldchangeinthegutandtheexpressionlevelsofonly2genesmeetthisthresholdintheMLN.Thisfairlysmallnumberofgenesmightnothavebeen164
spottedinthelargenumberofgeneswithsignificantchangesinexpressionlevels.
However,itmightproveusefultoincludegenesthatareup-ordown-regulatedbutdonotmeettheappliedthresholdtofindananswertothefollowingcomplexquestionatthenextleveloftheclassification(Fig.1)‘‘Whichpathwaysaredifferentlyactivatedinsusceptibleandresistantmice?’’
Asshowninthisanalysis,adifferentactivationofgenesinvolvedincaspasecascadeinapoptosiscanbeobserved.Applicationofthesameanalysistootherapoptosis-relatedpathways,suchasapoptoticsignallinginresponsetoDNAdamageorroleofmitochondriainapoptoticsignalling,showsasimilarpatternsuggestingadifferentactivationofapoptosispathwaysinthesusceptibleandresistantmousestrains.ThesefindingswillbecorroboratedexperimentallytodeterminethesignificanceforthedifferentoutcomesofinfectioninAKRandBALB/cmiceandtoanswerthefollowingbiologicalquestionfromthisanalysis(seeFig.1)‘‘Whichofthediffer-entlyactivatedpathwaysinsusceptible/resistantmicearesignificantindetermininghost-protectiveimmunity?’’
However,byrankingthegenesaccordingtotheirexpressionlevelsandapplyingathresholdtoexcludegeneswithnon-significantchangesinexpressionlevels,thisinformationcouldhavebeenmissed.Theclassificationcanthusbeusedtoidentifyapproachesthatdifferfromtheusualapproachofanalysingexperimentaldata.Thiscanmeananalys-ingexperimentaldatafromadifferentperspective,suchaspathwaysorfunctionalannotation,andcanrevealinformationthatmightotherwisehavebeenoverlooked.Thus,theapproachtakentoanalyseexperimentaldataisimportant.Exploringdatafromdifferentperspectivescanyieldnovelinformationandgeneratenewhypothesestobetestedexper-imentally.
Furthermore,theclassificationcanbeusedasananalysisofrequirementsforbioinformaticstoolsforimmunology.Itindicatesthekindsofanalysistasksthathavetobeprovidedtoallowuserstoanalysetheintegrateddatainabiologicallymeaningfulandcontext-richmanner.Applicationsforansweringsimplequestionsandtheircombinationscanbeimplementedquiteeasily,whereasinordertoanswerthemorecomplexquestions,sophisticatedanalysistechniquesarerequired.
TheclassificationhasbeenusedinthismannertoallowuserstoquerydifferentkindsofdataintegratedinthemouseGenomeInformationManagementSystem(GIMS)(Cornelletal.2003).Sofar,mostofthesimplequestionsandsomeoftheircombinationsareprovidedbythesystem.Thesystemwillbeextendedtoanswermorequestionsatdifferentlevelsofabstractionandcomplexitytoprovidethemeansforanalysingthestoreddatainasystematicway.
Aclassificationoftaskstostudyimmuneresponse
DISCUSSION
Wehavepresentedasystematicclassificationoftasksforimmunologicalbioinformaticsthatcanbeappliedtoanalyseexperimentaldata.Severaldifferentlevelsofquestionshavebeenidentifiedwhichareeitherdata-drivenordrivenbyimmunologicalknowledge.Thesearebasedonaclassificationofavailableandrelevantdatasourcesandofimmunologicalknowledge.Simpledata-driventaskscanbecombinedtoformmorecomplextasks,whichagaincanbecombinedtoanswerhigherlevelquestions.
Furthermore,wehaveshownwaystodeploythisclassification.Itcanbeusedforidentifyingdifferentwaystoanalyseandcombineavailabledata.Itcanalsobeusedtoidentifythequestionsthatneedtobeaskedandthetypesofdatathatneedtobeanalysedinordertoanswermoregeneralquestions.Thiswouldallowinsightstobegainedintotheimmunesystemwithitsrangeofavailableeffectormechanisms.BothwaysofdeployingtheclassificationhavebeenillustratedusingcasestudiesoftheimmuneresponseinmicetoinfectionwiththeintestinalnematodeparasitesT.murisandH.polygyrus.
Itisalsopossibletousetheclassificationasasetofrequirementstoguidethedevelopmentofdataanalysissoftwareforimmunology.Suchadisciplinedapproachcanprovidetheusersofthesoftwarewithstructuredfacilitiestoqueryandanalyseitsstoredcontentsinacontext-richandmeaningfulmanner.Severalofthesimplequestionsandtheircompo-sitionshavebeenimplementedinGIMS,whichhasinturnbeenusedtoexplorethecasestudiespresentedinthepaper.
Toevaluatetheusefulnessofourclassificationbeyondthetwocasestudiespresentedabove,wehavechosentoconsidersomerecentstudiesofhigh-throughputdata.Thesehavebeenchoseninthecontextofinfectionwitharangeofdiversepathogensandwehaveplacedtheanalysesundertakeninthesestudiesinourclassificationscheme.Eventhoughthesestudiesexaminedifferentaspectsofim-munology,mostofthemusesimilarapproachestoanalysegeneexpressiondata.Thestudiesincludetheresponseofthehosttoinfection(e.g.Domachowskeetal.2002;Jietal.2003;Cooketal.2004;Tongetal.2004),theinfectingagentinitseffortstoevadetheimmunesystemofthehost(e.g.Dahletal.2003),theinteractionbetweenhostandpathogen(e.g.Bladeretal.2001),andthereactionofanimmunizedhosttoinfection(e.g.Rahn,RedlineandBlanchard,2004;Byonetal.2005).
Thedatawereanalysedfromthebottombyidentifyingdifferentlyexpressedgenes,mainlyusingafold-changeapproach.Thiswasfollowedbytheidentificationofthefunctionalcharacteristicsofthesegenesorthepathwaysinwhichtheyareinvolved.Therefore,theanalysistasksusedatthe
165
firsttwolevelsoftheclassificationstartingfromthebottom(seeFig.1andsupplementarydatafile2)weremainlythefollowingquestions.Simplequestion,‘‘Whichgenesaredifferentlyregulated?’’Compositionofsimplequestions,‘‘Lookingatdifferentlyregulatedgenesandtheirfunctionalannotation,dotheyhavedifferentannotationsordotheyshareannotations?’’or‘‘Lookingatdiffer-entlyregulatedgenes,inwhichpathwaysaretheyinvolved?’’However,slightdifferencesintheanalysisapproachescanbeseen.Forexample,Domachowskeetal.(2002)focusedongeneswithaparticularfunction:inthiscasegenesinvolvedintheantiviralinflammatoryresponse.Cooketal.(2004),however,combinedmicroarrayanalysiswithQTLanalysistoidentifycandidategeneslocatedinQTLregionsthataredifferentlyexpressedinresistantandsusceptiblemice.Thesevariedapproachesarealsoseentofitwellwithinourclassificationscheme(seeFig.1andsupplementarydatafile2).
Basedontheaspectsofimmunologyexaminedinthesestudies,themoregeneralanalysescoverabroadrange.Theseincludeacomparisonofexpressionpatternsovertimepost-infection(e.g.Bladeretal.2001;Jietal.2003;Tongetal.2004)ortimepost-vaccinationofthehost(Byonetal.2005).Alsoincludedarecomparisonsofexpressionpatternsbetweeninfectionswithdifferentstrainsofpathogen(e.g.Dahletal.2003),betweendifferentpathogens(e.g.Bladeretal.2001),andbetweenimmunizedandnon-immunizedchallengedhosts(Rahnetal.2004).However,allofthesequestionsarepartoftheclassificationpresentedhere(seesupplementarydatafile2).Theyrepresentjustafewofthemanypossibleapproachestoanalysinghigh-throughputdataandcorrelatingitwithotheravailableinformation.Therefore,theclassificationisapplicabletostudyingmanydifferentaspectsofimmunitytoabroadrangeofpathogens.Itcanalsobeusedtoidentifymoreanalysistasksthatcanbecarriedoutontheavailabledataandcanhelptoexplorethedatamoresystem-aticallyandmorethoroughly.
Theclassificationbynomeanscontainsacompletelistofquestionsthatcanpossiblybeaskedtounlockthecomplexityoftheimmunesystem.Nordoesitprovideacompletelistofavailableandrelevantdatasources.However,tothebestoftheauthors’knowledgeitisthefirstattempttoclassifytasksanddatathatareofrelevancetoimmunologyinasystematicway.Webelievethatnewquestionsandkindsofdatathatwillarisewiththeadventofnewhigh-throughputtechniquescanbeplacedintotheexistingclassificationscheme.
TheauthorswouldliketothankAndyBrass,ChrisGarwood,PhilLord,andMikeCornellforvaluablecom-mentsontheclassificationandthemanuscript.ThisworkwassupportedbytheWellcomeTrust(grantreferencenumbers068639and044494).
C.Hedelerandothers
REFERENCES
Bader,G.D.,Betel,D.andHogue,C.W.V.(2003).BIND:theBiomolecularInteractionNetworkDatabase.NucleicAcidsResearch31,248–250.
Bancroft,A.J.,Else,K.J.andGrencis,R.K.(1994).Low-levelinfectionwithTrichurismurissignificantlyaffectsthepolarizationoftheCD4response.EuropeanJournalofImmunology24,3113–3118.
Bellaby,T.,Robinson,K.andWakelin,D.(1996).InductionofdifferentialT-Helper-CellresponsesinmiceinfectedwithvariantsoftheparasiticnematodeTrichurismuris.InfectionandImmunity64,791–795.Blader,I.J.,Manger,I.andBoothroyd,J.C.(2001).MicroarrayAnalysisRevealsPreviouslyUnknownChangesinToxoplasmagondii-infectedHumanCells.TheJournalofBiologicalChemistry276,24223–24231.Blake,J.A.,Richardson,J.E.,Bult,C.J.,Kadin,J.A.,Eppig,J.T.andthemembersoftheMouse
GenomeDatabaseGroup(2003).MGD:TheMouseGenomeDatabase.NucleicAcidsResearch31,193–195.Bretscher,P.A.,Wei,G.,Menon,J.N.and
Bielefeldt-Ohmann,H.(1992).Establishmentofstable,cell-mediatedimmunitythatmakes
‘‘susceptible’’miceresistanttoLeishmaniamajor.Science5069,539–542.
Byon,J.Y.,Ohira,T.,Hirono,I.andAoki,T.(2005).UseofacDNAmicroarraytostudyimmunityagainstviralhemorrhagicsepticaemia(VHS)inJapaneseflounder(Paralichthysolivaceus)followingDNA
vaccination.FishandShellfishImmunology18,135–147.
Bystro
¨m,J.,Wynn,T.A.,Domachowske,J.B.andRosenberg,H.F.(2004).Genemicroarrayanalysisrevealsinterleukin-5-dependenttranscriptionaltargetsinmousebonemarrow.Blood103,868–877.
Cornell,M.,Paton,N.W.,Hedeler,C.,Kirby,P.,Delneri,D.,Hayes,A.andOliver,S.G.(2003).GIMS:anintegrateddatastorageandanalysis
environmentforgenomicandfunctionaldata.Yeast20,1291–1306.
Cook,D.N.,Wang,S.,Wang,Y.,Howles,G.P.,Whitehead,G.S.,Berman,K.G.,Church,T.D.,Frank,B.C.,Gaspard,R.M.,Yu,Y.,
Quackenbush,J.andSchwarz,D.A.(2004).Geneticregulationofendotoxin-inducedairwaydisease.Genomics83,961–969.
Croker,B.A.,Krebs,D.l.,Zhang,J.G.,Wormald,S.,Willson,T.A.,Stanley,E.G.,Robb,l.,
Greenhalgh,C.J.,Fo
¨rster,I.,Clausen,B.E.,Nicola,N.A.,Metcalf,D.,Hilton,D.J.,
Roberts,A.W.andAlexander,W.S.(2003).SOCS3negativelyregulatesIL-6signalinginvivo.NatureImmunology4,540–545.
Dahl,J.L.,Kraus,C.N.,Boshoff,H.I.M.,Doan,B.,Foley,K.,Avarbock,D.,Kaplan,G.,Mizrahi,V.,Rubin,H.andBarryIII,C.E.(2003).TheroleofRelMtb-mediatedadaptationtostationaryphaseinlong-termpersistenceofMycobacteriumtuberculosisinmice.ProceedingsoftheNationalAcademyofSciences,USA100,10026–10031.
Deschoolmeester,M.L.andElse,K.J.(2002).
CytokineandchemokineresponsesunderlyingacuteandchronicTrichurismurisinfection.InternationalReviewsofImmunology21,439–467.
166
Domachowske,J.B.,Bonville,C.A.,Easton,A.J.andRosenberg,H.F.(2002).Differentialexpressionofproinflammatorycytokinegenesinvivoinresponsetopathogenicandnonpathogenic
pneumovirusinfections.TheJournalofInfectiousDiseases186,8–14.
Edwards,A.D.,Chaussabel,D.,Tomlinson,S.,Schulz,O.,Sher,A.andReisE.SousaC.(2003).RelationshipsamongmurineCD11c(high)dendriticcellsubsetsasrevealedbybaselinegeneexpressionpatterns.JournalofImmunology171,47–60.
Else,K.J.andWakelin,D.(1988).TheeffectsofH-2andnon-H-2genesontheexpulsionofthenematodeTrichurismurisfrominbredandcongenicmice.Parasitology96,543–550.
Eppig,J.,Blake,J.A.,Burkart,D.,Goldsmith,C.,Lutz,C.andSmith,C.(2002).Corrallingconditionalmutations:aunifiedresourceformousephenotypes.Genesis32,63–65.
Gollub,J.,Ball,C.A.,Binkley,G.,Demeter,J.,Finkelstein,D.B.,Hebert,J.M.,Hernandez-boussard,T.,Jin,H.,Kaloper,M.,Matese,J.C.,Schroeder,M.,Brown,P.O.,Botstein,D.andSherlock,G.(2003).TheStanfordMicroarrayDatabase:dataaccessandqualityassessmenttools.NucleicAcidsResearch31,94–96.
Grencis,R.K.(2001).Cytokineregulationofresistanceandsusceptibilitytointestinalnematode
infection–fromhosttoparasite.VeterinaryParasitology100,45–49.
Hoffmann,K.F.,McCarty,T.C.,Segal,D.H.,Chiaramonte,M.,Hesse,M.,Davis,E.M.,
Cheever,A.W.,Meltzer,P.S.,Morse,H.C.3rdandWynn,T.A.(2001).DiseasefingerprintingwithcDNAmicroarraysrevealsdistinctgeneexpressionprofilesinlethaltype-1andtype-2cytokine-mediated
inflammatoryreactions.TheFASEBJournal15,2545–2547.
Hubbard,T.,Barker,D.,Birney,E.,Cameron,G.,Chen,Y.,Clark,L.,Cox,T.,Cuff,J.,Curwen,V.,Down,T.,Durbin,R.,Eyras,E.,Gilbert,J.,Hammond,M.,Huminiecki,L.,Kasprzyk,A.,Lehvaslaiho,H.,Lijnzaad,P.,Melsopp,C.,Mongin,E.,Pettett,R.,Pocock,M.,Potter,S.,
Rust,A.,Schmidt,E.,Searle,S.,Slater,G.,Smith,J.,Spooner,W.,Stabenau,A.,Stalker,J.,Stupka,E.,Ureta-Vidal,A.,Vastrik,I.andClamp,M.(2002).TheEnsemblgenomedatabaseproject.NucleicAcidsResearch30,38–41.
Ji,M.J.,Su,C.,Wu,H.W.,Zhu,X.,Cai,X.P.,Li,C.L.,Li,G.F.,Wang,Y.,Zhang,Z.S.andWu,G.L.
(2003).GeneexpressionprofileofCD4+TcellsrevealsaninterferonsignalingsuppressionassociatedwithprogressionofexperimentalSchistosomajaponicuminfection.CellularImmunology224,55–61.Kanehisa,M.,Goto,S.,Kawashima,S.andNakaya,A.(2002).TheKEGGdatabasesatGenomeNet.NucleicAcidsResearch30,42–46.
Kell,D.B.andOliver,S.G.(2003).Hereistheevidence,nowwhatisthehypothesis?Thecomplementaryrolesofinductiveandhypothesis-drivenscienceinthepost-genomicera.BioEssays26,99–105.
Lang,R.,Pauleau,A.L.,Parganas,E.,Takahashi,Y.,Mages,J.,Ihle,J.N.,Rutschman,R.and
Aclassificationoftaskstostudyimmuneresponse
Murray,P.J.(2003).SOCS3regulatestheplasticityofgp130signaling.NatureImmunology4,546–550.Mak,T.W.,Penninger,J.M.andOhashi,P.S.(2001).KnockoutMice:AParadigmShiftin
ModernImmunology.NatureReviewsImmunology1,11–19.
Menge,D.M.,Behnke,J.M.,Lowe,A.,Gibson,J.P.,Iraqi,F.A.,Baker,R.L.andWakelin,D.(2003).Mappingofchromosomalregionsinfluencingimmunologicalresponsestogastrointestinal
nematodeinfectionsinmice.ParasiteImmunology25,314–349.
Mueller,A.,O’Rourke,J.,Grimm,J.,Guillemin,K.,Dixon,M.F.,Lee,A.andFalkow,S.(2003).Distinctgeneexpressionprofilescharacterizethe
histopathologicalstagesofdiseaseinHelicobacter-inducedmucosa-associatedlymphoidtissuelymphoma.ProceedingsoftheNationalAcademyofSciences,USA100,1292–1297.
Mulder,N.,Apweiler,R.,Attwood,T.,Bairoch,A.,Barrell,D.,Bateman,A.,Binns,D.,Biswas,M.,Bradley,P.,Bork,P.,Bucher,P.,Copley,R.,Courcelle,E.,Das,U.,Durbin,R.,Falquet,L.,Fleischmann,W.,Griffiths-Jones,S.,Haft,D.,Harte,N.,Hulo,N.,Kahn,D.,Kanapin,A.,Krestyaninova,M.,Lopez,R.,Letunic,I.,Lonsdale,D.,Silventoinen,V.,Orchard,S.,Pagni,M.,Peyruc,D.,Ponting,C.,Selengut,J.,Servant,F.,Sigrist,C.,Vaughan,R.andZdobnov,E.(2003).TheInterProDatabase,2003bringsincreasedcoverageandnewfeatures.NucleicAcidsResearch31,315–318.
Noordewier,M.andWarren,P.V.(2001).Geneexpressionmicroarraysandtheintegrationof
biologicalknowledge.TRENDSinBiotechnology19,412–415.
Pan,W.(2002).Acomparativereviewofstatistical
methodsfordiscoveringdifferentiallyexpressedgenesin
167
replicatedmicroarrayexperiments.Bioinformatics18,546–554.
Quackenbush,J.(2002).Microarraydatanormalizationandtransformation.NatureGenetics32(Suppl.),496–501.
Rahn,W.,Redline,R.W.andBlanchard,T.G.(2004).MolecularanalysisofHelicobacterpylori-associatedgastricinflammationinnaı¨veversusimmunizedmice.Vaccine23,807–818.
Ricciardi-Castagnoli,P.andGranucci,F.(2002).Interpretationofthecomplexityofinnateimmuneresponsesbyfunctionalgenomics.NatureReviewsImmunology2,1–8.
Rogner,U.C.andAvner,P.(2003).Congenicmice:cuttingtoolsforcompleximmunedisorders.NatureReviewsImmunology3,243–251.
Sherlock,G.(2000).Analysisoflarge-scalegeneexpressiondata.CurrentOpinioninImmunology12,201–205.
TheGeneOntologyConsortium(2000).Gene
Ontology:toolfortheunificationofbiology.NatureGenetics25,25–29.
Tong,H.H.,Long,J.P.,Li,D.andDeMaria,T.F.(2004).AlterationofgeneexpressioninhumanmiddleearepithelialcellsinducedbyinfluenzaAvirusanditsimplicationforthepathogenesisofotitismedia.MicrobialPathogenesis37,193–204.
Wheeler,D.L.,Church,D.M.,Federhen,S.,Lash,A.E.,Madden,T.L.,Pontius,J.U.,Schuler,G.D.,Schriml,L.M.,Sequeira,E.,Tatusova,T.A.andWagner,L.(2003).DatabaseresourcesoftheNationalCenterforBiotechnology.NucleicAcidsResearch31,28–33.
Xenarios,I.,Salwinski,L.,Duan,X.J.,Higney,P.,Kim,S.M.andEisenberg,D.(2002).DIP,theDatabaseofInteractingProteins:aresearchtoolforstudyingcellularnetworksofproteininteractions.NucleicAcidsResearch30,303–305.
因篇幅问题不能全部显示，请点此查看更多更全内容