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.
Comparativegenemarkerselectionsuite
Fig.1.PlotofteststatisticscoreandtableofsignificancevaluesfordatainGolubetal.(1999).
q-value.Theq-valueattemptstogainitsextrapowerfromesti-matingp0(0 p0 1),thefractionoftruenullhypothesesamongalltestedhypotheses(theBHprocedurealwaysassumesp0¼1).Whenthisfractionislarge,whichisthecaseformanymicroarrayexperimentsinwhichwetestthousandsofgenesofwhichweexpectveryfewtobedifferentiallyexpressed,littleadvantageisgainedbyusingtheq-value.However,ifweexpectalargeproportionofthetestedgenestobedifferentiallyexpressed,theq-valuemightallowforthereductionoffalsenegativeswhencomparedwiththeBHmethod.Weincludeaplotofthep0estimateversusthetuningparameterltoevaluatetheaccuracyofthe nalp0estimate.Whenamoreconservativeapproachisrequired,wesuggestcontrollingthefamily-wiseerrorrate(FWER),theprobabilityofhavingatleastonefalsepositive.Forexample,theFWERmaybepreferredwhenfurtherinvestigationofanyfalsepositiveiscostly.WeincludethreemethodsforcontrollingtheFWER.TheBonfer-ronimethodisthemostconservative,followedbytheempiricalFWERandthemaxTprocedure(WestfallandYoung,1993).Sometimesdatacontainextraneousvariablesthatarenotaccountedforinthedesignoftheexperimentandthatcandistorttheresults.Forexample,selectingmarkersthatdistinguishtumorfromnon-tumorsamplesmightleadtoincorrectresultsifsomeofthesamplesaremaleandsomearefemale.Weprovidetheoptiontocontrolfortheseconfoundingphenotypesbyprovidingarestrictedpermutations(Good,1994;Lu,2005)option,inwhichtheclasslabelsareshuf edonlywithineachconfoundingphenotype.Intheexampleinwhichgenderistheconfoundingphenotype,wecanrestrictthepermutationssothatfemalenon-tumorlabelsareonlypermutedwithfemaletumorlabelsandmalenon-tumorlabelsareonlyshuf edwithmaletumorlabels.
hypothesisrejectioncriteria.Userscanvisuallyinspectthepro lesofeachfeatureacrosseachsampleinaheatmaporexpressionpro leformat.
Userscanviewfeaturesthatpassselected lteringcriteriaandcreatederivativedatasetsandfeatureslistsfromthese lteredfea-tures.Forexample,ausercanextractallgenesthathaveaq-value<0.1andsavethecorrespondingdatasetandmarkerlist.Weincludeautilitymoduletoautomatethisfunction.Alltheplotsinthevieweraredynamicallyupdatedtoincludeonlythefeaturesthatpasstheselected lteringcriteria.
Theannotationoffeaturesisprovidedbytwomechanisms.Affymetrixprobesetidenti erscanbeinteractivelyannotatedfromgenomicdatabasessuchasGenBank,UniGene,SwissProt,LocusLinkandGeneOntologyusingtheGeneCruiserWebservice(Liefeldetal.,2005).Userscanalsoentertheirownannotationsoffeaturesandviewtheseannotationsviaacolor-codingmechanism.
ACKNOWLEDGEMENTS
TheauthorswishtothankthefollowingmembersoftheCancerProgramattheBroadInstitute:ToddGolub,TedLiefeldandDavidTwomey.GenePatternissupportedbyfundingfromtheNationalInstitutesofHealth.ConflictofInterest:nonedeclared.
REFERENCES
Benjamini,Y.andHochberg,Y.(1995)Controllingthefalsediscoveryrate:apractical
andpowerfulapproachtomultipletesting.J.RStat.Soc.B(Methodological),57,289–300.
Golub,T.etal.(1999)MolecularClassi cationofCancer:classDiscoveryandClass
PredictionbyGeneExpression.Science,286,531–537.
Good,P.(1994)PermutationTests:APracticalGuideforTestingHypotheses.
Springer-Verlag,NY.
Liefeld,T.etal.(2005)GeneCruiser:awebservicefortheannotationofmicroarray
data.Bioinformatics,21,3681–3682.
Lu,J.etal.(2005)MicroRNAExpressionPro lesClassifyHumanCancers.Nature,
435,834–838.
Reich,M.etal.(2006)GenePattern2.0.NatureGenetics,38,500-501.
Storey,J.D.andTibshirani,R.(2003)Statisticalsigni canceforgenomewidestudies.
PNAS,100,9440–9445.
Westfall,P.H.andYoung,S.S.(1993)Resampling-basedmultipletesting:examples
andmethodsforP-valueadjustment.WileySeriesinProbabilityandStatistics.-Wiley,

NY.
VISUALIZATIONANDUTILITYMODULES
Thevisualizationportionofthesuiteprovidesaframeworkforassessingtheresultsfromtheanalyticmodule.Themoduleincludesinteractivehistogramsusedtodeterminenulldistributionsforeachmeasureofsigni cance.Additionallyaplotoftheteststatisticrankversustheteststatisticisincluded,whichisusefulforvisualizingthenumberoffeaturesthatareupregulatedineachclass(Figure1).Pairwisecomparisonofdifferentsigni cancemeasurescanbeplottedtohelpusersassesstherelativestringencyoftheselected
1925
BIOINFORMATICS APPLICATIONS NOTE Gene expression Vol. 23 no. 3 2007, pages 387–389 doi:10.1093/bioinformatics/btl600 Mosclust: a software library for ...
BIOINFORMATICS APPLICATIONS NOTE Gene expression MAGIC Tool integrated microarray data anal_专业资料。doi:10.1093/bioinformatics/bti247...
BIOINFORMATICS APPLICATIONS NOTE Gene expression Bloader—a batch loader application for MI Summary: BLoader is a client server application for annotating and...
BIOINFORMATICS APPLICATIONS NOTE doi10.1093bioinformaticsbtk005 Gene expression...cance Testing, with Applications to Comparative Microarray Experiments. UW ...
BIOINFORMATICS APPLICATIONS NOTE Gene expression Vol. 23 no. 13 2007, ...Pan,W. (2002) A comparative review of statistical methods for discovering ...
automating and visualizing comparative genome mapsVol... BIOINFORMATICS APPLICATIONS NOTE Genome analysis Vol...cation of CS, CSO, breakpoints, marker/gene ...
BIOINFORMATICS APPLICATIONS NOTE Gene expression Vol. 23 no. 9 2007, pages 1168–1169 doi:10.1093/bioinformatics/btm072 Codelink: an R package for ...
expression profilesexpression profiles隐藏>> BIOINFORMATICS APPLICATIONS NOTE Gene expression Vol. 22 no. 2 2006, pages 245–247 doi:10.1093/bioinformatics/...
BIOINFORMATICS APPLICATIONS NOTE Gene expression Vol. 23 no. 9 2007, pages 1168–1169 doi:10.1093/bioinformatics/btm072 Codelink: an R package for ...
热门文档