文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > BIOINFORMATICS APPLICATIONS NOTE Gene expression Comparative gene marker selection suite
侵权投诉

BIOINFORMATICS APPLICATIONS NOTE Gene expression Comparative gene marker selection suite

Motivation: An important step in analyzing expression profiles from microarray data is to identify genes that can discriminate between distinct classes of samples. Many statistical approaches for assigning significance values to genes have been developed.

BIOINFORMATICSAPPLICATIONSNOTE

Geneexpression

Vol.22no.152006,pages1924–1925

doi:10.1093/bioinformatics/btl196

Comparativegenemarkerselectionsuite

JoshuaGouldÃ,GadGetz,StefanoMonti,MichaelReichandJillP.Mesirov

BroadInstituteofMITandHarvard,Cambridge,MA02142,USA

ReceivedonMarch13,2006;revisedonApril21,2006;acceptedonMay15,2006AdvanceAccesspublicationMay18,2006AssociateEditor:NikolausRajewsky

ABSTRACT

Motivation:Animportantstepinanalyzingexpressionprofilesfrommicroarraydataistoidentifygenesthatcandiscriminatebetweendistinctclassesofsamples.Manystatisticalapproachesforassigningsignificancevaluestogeneshavebeendeveloped.TheComparativeMarkerSelectionsuiteconsistsofthreemodulesthatallowuserstoapplyandcomparedifferentmethodsofcomputingsignificanceforeachmarkergene,aviewertoassesstheresults,andatooltocreatederivativedatasetsandmarkerlistsbasedonuser-definedsignificancecriteria.

Availability:TheComparativeMarkerSelectionapplicationsuiteisfreelyavailableasaGenePatternmodule.TheGenePatternanalysisenvironmentisfreelyavailableatdoc.guandang.net

Contact:jgould@broad.mit.edu

1INTRODUCTION

Whenanalyzinggenome-widetranscriptionpro lesfrommicroar-raydata,the rststepisoftentoidentifygenesthatcandiscriminatebetweendistinctclassesofsamples(usuallyde nedbyaphenotype,suchastumorornormal).Thisprocessiscommonlyreferredtoasmarker(orfeature)selection.Manystatisticalapproacheshavebeendevelopedtoestimatethesigni canceofmarkergenes.TheComparativeMarkerSelectionsuiteimplementsmanyoftheseapproaches,andprovidesvisualizationtoolsforeasycomparisonoftheresultsthataregenerated.ThesuiteconsistsofthreemodulesintheGenePattern(Reichetal.,2006)analysisenvironmentandincludes(1)ananalyticmodulethatcomputesthestatisticalsig-ni canceofeachgene,andincludesseveralmethodsofcorrectingformultiplehypothesistesting(MHT),(2)avisualizationmoduletoaidintheevaluationoftheanalyticmoduleresults,and(3)autilitymoduletocreatederivativedatasetsandmarkerlistsbasedonuser-de nedsigni cancecriteria.CompletedocumentationforGenePatternisprovidedontheGenePatternwebsite.Additionally,eachmoduleintheGenePatternenvironmentcontainsdetaileddocumentation.Wealsoprovidedefaultsettingsforallmoduleparameters.

2ANALYTICMODULE

Theanalyticmoduletakesasinputadatasetofexpressionpro lesfromsamplesbelongingtotwophenotypes.Eithertwo-channel(cDNA)orabsolutevalue(Affymetrix)datacanbeusedasinputtothemoduleandanymissingvaluescanbeimputed(seeFAQon

Ã

Towhomcorrespondenceshouldbeaddressed.

theGenePatternwebsite).Ifadatasetcontainsmultiplephenotypes,thenthereistheoptiontoperformallpairwisecomparisonsorallone-versus-allcomparisons.Ateststatistic(e.g.t-test)ischosentoassessthedifferentialexpressionbetweenthetwoclassesofsamples.Notethattechnicalandbiologicalreplicatesarehandledthesamewayasindependentsamples.Thesigni cance(nominalP-value)ofmarkergenesiscomputedusingapermutationtest,whichisacommonlyusedmethodforassessingthesigni canceofmarkergenes.Permutationtestshavetheadvantageofnotassumingaparametricunderlyingdistributionofexpressionvalues,andimportantlytheypreservegene–genecorrelationswhichaffectsomemeasurementsofsigni cance.Toconstructadistributionoftheteststatistic,underthenullhypothesisofnodifferentialexpres-sion,phenotypelabelsarerandomlyre-assignedtosamplesandtheteststatisticisrecomputedfortherelabeleddataset.Thisprocedureisrepeatedforagivennumberofrelabelingstoyieldtheempiricalnull.Itshouldbenotedthatthetotalnumberofpossibleexhaustivepermutationsisafunctionofthenumberofsamplesineachclass(forexamplegiven20samples,10ofeachclass,thetotalnumberofdistinctpermutationsisð20Þ.Theminimumnumberofpermuta-10

tionsnecessaryisafunctionofthenumberofhypothesestestedandthenumberofrejectedhypothesesweexpect.Inmostcases,wesuggestaminimumof10samplesperclass.Whenthetotalnumberofpermutationsisnotlargeenoughtoestimateasuf cientlyaccu-rateP-value,themoduleprovidestheoptionofcomputingasymp-toticP-valuesbasedonthet-test.InadditiontoreportingthenominalP-value,wealsoreporttheestimated95%con denceintervalsforthenominalP-valuetoassessP-valueaccuracy.WealsoincludeanoptiontoperformallpossiblerelabelingstoobtainexactP-values.

SelectingclassmarkersisaparticularinstanceofthegeneralMHTproblem.Sinceseveralthousandhypothesesareusuallytestedatonce,thenominalP-valueshavetobecorrectedtoaccountfortheincreasednumberofpotentialfalsepositives.Forexample,ifwetest20000genesfordifferentialexpression,anominalP-valuethresholdof0.01wouldonlyensurethattheexpectednumberoffalsepositivesis<200(0.01Ã20000).

OneapproachtoadjustingforMHTistocontrolthefalsedis-coveryrate(FDR),theexpectedfractionoffalsepositivesamongallgenesreportedassigni cant.Inmostcases,controllingtheFDRissuf cientbecausethepurposeofmostmicroarrayinvestigationsistogeneratehypothesesforfurtherstudy.TheFDRcut-offlevelcontrolsthefractionoffalseleadsthattheuseriswillingtotolerate.WeincludetwomethodsforcomputingtheFDR:theBHproceduredevelopedbyBenjaminiandHochberg(1995)andtheq-valuemethodofStoreyandTibshirani(2003).TheBHpro-ceduregivesamoreconservativeestimateoftheFDRthanthe

1924

ÓTheAuthor2006.PublishedbyOxfordUniversityPress.Allrightsreserved.ForPermissions,pleaseemail:journals.permissions@oxfordjournals.org

第1页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email