30360.基于boosting策略的啟動子預測方法研究

1、煙臺大學碩士學位論文基于Boosting策略的啟動子預測方法研究姓名：曾慶尚申請學位級別：碩士專業(yè)：計算機應用技術(shù)指導教師：武栓虎20090331IIAbstractComputationalpredictionofeukaryoticpromoterisoneofmostchallengingproblemsinDNAsequenceanalysis.Astranionisthefirststepofgeneexpressionreg

2、ulationontranionisanimptantwayofexpressionregulationaswellpromoterdeterminesthepositionoftranionalstartpointthefrequencywithwhichthegeneistranscribedsopromoterrecognitionisimptantfgenomeannotation.Basedontheknownpromoter

3、nonpromotersequencestwonewdifferentmethodsarepresentedfpromoterprediction.FirstofallanewmethodofpromoterpredictioncalledKLBoostingisproposedbythestudyitisbasedonthefollowinghypothesis:Promoterisdeterminedbysomewdpatterns

4、differentpromotersaredeterminedbydifferentwds.Mostpotentialfeaturesareedbydivergencedistancetobuildasequenceofweakclassifierbyfeatureboosting.Anumberofweakclassifiersconstructastrongclassifierwhichcanachieveabetterperfma

5、nce.Differentfromotherclassifiersadifferenttrainingclassifyingstrategyisadopted.SecondlyanotherpromoterpredictionmethodcalledPWMBoostingispresentedinthestudy.itisbasedonthefollowinghypothesis:Promoterisdeterminednotonlyb

6、ysomewdpatternsbutalsobythepositionsofwdpatterns.PotentialfeaturesofthesequenceareedbysearchingthediversityofpositionweightmatrixtheyaretrainedpredictedbycallingtheboostingalgithmoftheOpenSourceComputerVisionLibrary.Itad

7、optsasimilartrainingclassifyingstrategywithKLBoostingmethod.ExperimentalresultsontestsamplesshowthatKLBoostingmethodhasamoderatesensitivityspecificityinadditionitsstabilityalsoisbetter.Althoughthesensitivitystabilityofth

8、ePWMBoostingmethodisnotbetterthantheKLBoostingmethoditsspecificityishigherthanKLBoosting.Finallyexperimentalresultsofthetwoalgithmsinthestudyonlargegenomicsequencescomparingwithotherfourexcellentalgithms(EponinePDFFirstE

9、FPromoterInspect)showtheyareefficientwithhighersensitivityspecificityinpredictingpromoterregions，havetheirownacteristics.KeyWds:DNAsequenceanalysisdivergencedistancefeatureboostingpositionweightmatrixOpenSourceComputerVi