資源描述：
《gammaMAXT a fast multiple-testingcorrection algorithmGAMAMAXT：快速多重測(cè)試 校正算法》由會(huì)員上傳分享，免費(fèi)在線閱讀，更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫(kù)。
1、VanLishoutetal.BioDataMining(2015)8:36DOI10.1186/s13040-015-0069-xBioDataMiningSOFTWAREARTICLEOpenAccessgammaMAXT:afastmultiple-testingcorrectionalgorithmFran?oisVanLishout1,2*,FrancescoGadaleta1,2,JasonH.Moore3,LouisWehenkel1,2andKristelVanSteen1,2*Correspondence:f
2、.vanlishout@ulg.ac.beAbstract1SystemsandModelingUnit,Background:ThepurposeoftheMaxTalgorithmistoprovideasignificancetestMontefioreInstitute,Universityofalgorithmthatcontrolsthefamily-wiseerrorrate(FWER)duringsimultaneousLiège,Alléedeladécouverte10,4000Liège,Belgiumh
3、ypothesistesting.However,therequirementsintermsofcomputingtimeand2BioinformaticsandModeling,memoryofthisprocedureareproportionaltothenumberofinvestigatedhypotheses.GIGA-R,Avenuedel’H?pital1,4000Thememoryissuehasbeensolvedin2013byVanLishout’simplementationofMaxT,Sart
4、-Tilman,BelgiumFulllistofauthorinformationiswhichmakesthememoryusageindependentfromthesizeofthedataset.ThisavailableattheendofthearticlealgorithmisimplementedinMBMDR-3.0.3,asoftwarethatisabletoidentifygeneticinteractions,foravarietyofSNP-SNPbasedepistasismodelseffec
5、tively.Ontheotherhand,thatimplementationturnedouttobelesssuitableforgenome-wideinteractionanalysisstudies,duetotheprohibitivecomputationalburden.Results:InthisworkweintroducegammaMAXT,anovelimplementationofthemaxTalgorithmformultipletestingcorrection.Thealgorithmwas
6、implementedinsoftwareMBMDR-4.2.2,aspartoftheMB-MDRframeworktoscreenforSNP-SNP,SNP-environmentorSNP-SNP-environmentinteractionsatagenome-widelevel.Weshowthat,intheabsenceofinteractioneffects,test-statisticsproducedbytheMB-MDRmethodologyfollowamixturedistributionwitha
7、pointmassatzeroandashiftedgammadistributionforthetop10%ofthestrictlypositivevalues.WeshowthatthegammaMAXTalgorithmhasapowercomparabletoMaxTandmaintainsFWER,butrequireslesscomputationalresourcesandtime.Weanalyzeadatasetcomposedof106SNPsand1000individualswithinonedayo
8、na256-corecomputercluster.Thesameanalysiswouldtakeabout104timeslongerwithMBMDR-3.0.3.Conclusions:TheseresultsarepromisingforfutureGWAIs.Ho
