生物信息的论文文章
BIOINFORMATICSVol.17no.62001
Pages

520–525
MissingvalueestimationmethodsforDNA
microarrays
OlgaTroyanskaya1,MichaelCantor1,GavinSherlock2,
PatBrown3,TrevorHastie4,RobertTibshirani4,DavidBotstein2
andRussB.Altman1,
1StanfordMedicalInformatics,2DepartmentofGenetics,StanfordUniversitySchool
ofMedicine,Stanford,CA,USA,3DepartmentofBiochemistry,StanfordUniversity
SchoolofMedicine,andHowardHughesMedicalInstitute,Stanford,CA,USAand4DepartmentsofStatisticsandHealthResearchandPolicy,StanfordUniversity,
Stanford,CA,USA
ReceivedonNovember13,2000;revisedonFebruary22,2001;acceptedonFebruary26,2001
ABSTRACTINTRODUCTION
Motivation:Geneexpressionmicroarrayexperimentscan
generatedatasetswithmultiplemissingexpressionval-DNAmicroarraytechnologyallowsforthemonitoringues.Unfortunately,manyalgorithmsforgeneexpressionofexpressionlevelsofthousandsofgenesunderaanalysisrequireacompletematrixofgenearrayvaluesasvarietyofconditions(DeRisietal.,1997;Spellmaninput.Forexample,methodssuchashierarchicalcluster-etal.,1998).MicroarrayshavebeenusedtostudyaingandK-meansclusteringarenotrobusttomissingdata,varietyofbiologicalprocesses,fromdifferentialgeneandmayloseeffectivenessevenwithafewmissingvalues.expressioninhumantumors(Perouetal.,2000)toyeastMethodsforimputingmissingdataareneeded,therefore,sporulation(Chuetal.,1998).Variousanalysistechniquestominimizetheeffectofincompletedatasetsonanaly-havebeendeveloped,aimedprimarilyatidentifyingses,andtoincreasetherangeofdatasetstowhichtheseregulatorypatternsorsimilaritiesinexpressionunderalgorithmscanbeapplied.Inthisreport,weinvestigatesimilarconditions.Commonlyusedanalysismethodsautomatedmethodsforestimatingmissingdata.includeclusteringtechniques(Eisenetal.,1998;TamayoResults:Wepresentacomparativestudyofseveraletal.,1999),techniquesbasedonpartitioningofdatamethodsfortheestimationofmissingvaluesingene(Heyeretal.,1999;Tamayoetal.,1999),aswellasmicroarraydata.Weimplementedandevaluatedthreevarioussupervisedlearningalgorithms(Alteretal.,2000;methods:aSingularValueDecomposition(SVD)basedBrownetal.,2000;Golubetal.,1999;Raychaudhurietmethod(SVDimpute),weightedK-nearestneighbors(KN-al.,2000;Hastieetal.,2000).
Nimpute),androwaverage.WeevaluatedthemethodsThedatafrommicroarrayexperimentsisusuallyinusingavarietyofparametersettingsandoverdifferentrealtheformoflargematricesofexpressionlevelsofgenesdatasets,andassessedtherobustnessoftheimputation(rows)underdifferentexperimentalconditions(columns)methodstotheamountofmissingdataovertherangeofandfrequentlywithsomevaluesmissing.Missingvalues1–20%missingvalues.WeshowthatKNNimputeappearsoccurfordiversereasons,includinginsuf cientresolution,toprovideamorerobustandsensitivemethodformissingimagecorruption,orsimplyduetodustorscratchesonvalueestimationthanSVDimpute,andbothSVDimputetheslide.MissingdatamayalsooccursystematicallyandKNNimputesurpassthecommonlyusedrowaverageasaresultoftheroboticmethodsusedtocreatethem.method(aswellas llingmissingvalueswithzeros).WeOurinformalanalysisofthedistributionofmissingreportresultsofthecomparativeexperimentsandprovidedatainrealsamplesshowsacombinationofallofrecommendationsandtoolsforaccurateestimationofthese,butnonedominating.Suchsuspiciousdataismissingmicroarraydataunderavarietyofconditions.usuallymanually aggedandexcludedfromsubsequentAvailability:Thesoftwareisavailableatdoc.guandang.net(Alizadehetal.,2000).Manyanalysismethods,stanford.edu/projects/helix/pubs/impute/suchasprinciplecomponentsanalysisorsingularvalueContact:russ.altman@stanford.edudecomposition,requirecompletematrices(Alteretal.,
2000;Raychaudhurietal.,2000).Ofcourse,onesolution
Towhomcorrespondenceshouldbeaddressed.tothemissingdataproblemistorepeattheexperiment.
Thisstrategycanbeexpensive,buthasbeenusedin
520 cOxfordUniversityPress2001Downloaded from bioinformatics.oxfordjournals.org by guest on January 5, 2011
Unsupervised learning with non-ignorable missing data_专业资料。In this paper we explore the topic of unsupervised learning in the presence of nonignorable ...
Appl Intell DOI 10.1007s10489-006-0032-0 Semi-parametric optimization for missing data impu Abstract Missing data imputation is an important issue in ...
Missing data and the accuracy of Bayesian phylogenetics_专业资料 暂无评价
missing data, analysis of 10 000 s of molecular traits, permutation for ...Downloaded from http://bioinformatics.oxfordjournals.org/ at Sun Yat-Sen ...
How to avoid missing data and the problems they pose: design considerations_专业资料。海精神医学 21 年第 2 卷第 3024期 ?181 ? ?Botttsnpyhar9 ? ...
Robust probabilistic PCA with missing data and contribution analysis for outlier detection_专业资料。Robust probabilistic PCA with missing data and contribution...
LETTER Communicated by Erkki Oja Variational Bayesian Learning of ICA with Missing Data Missing data are common in real-world data sets and are a ...
SPSS for Windows操作(第三版)_计算机硬件及网络_IT/计算机_专业资料。中南...不同的变量,行列相交的单 元格中未能记录应有的数据即为缺失值(missing data)...
doi:10.1093/bioinformatics/bti345 Collateral missing value imputation: a new robust missing value estimation algorithm for microarray data Muhammad Shoaib B....
26 EM Algorithm 27 Missing Data ? We think of clustering as a problem of estimating missing data. ? The missing data are the cluster labels. ? ...
热门文档