Machine Learning Algorithms机器学习算法

《Machine Learning Algorithms机器学习算法》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、GCATTACGgenesGCATArticleIdentificationofDifferentiallyExpressedGenesbetweenOriginalBreastCancerandXenograftUsingMachineLearningAlgorithmsDelingWang1,2,†,Jia-RuiLi3,†,Yu-HangZhang1,LeiChen4ID,TaoHuang1,*IDandYu-DongCai3,*ID1InstituteofHealthSciences,ShanghaiInstitutesforBio

2、logicalSciences,ChineseAcademyofSciences,Shanghai200031,China;wangdl@sysucc.org.cn(D.W.);zhangyh825@163.com(Y.-H.Z.)2DepartmentofMedicalImaging,SunYat-senUniversityCancerCenter,StateKeyLaboratoryofOncologyinSouthChina;CollaborativeInnovationCenterforCancerMedicine,Guangzh

3、ou510060,China3SchoolofLifeSciences,ShanghaiUniversity,Shanghai200444,China;jiaruili@shu.edu.cn4CollegeofInformationEngineering,ShanghaiMaritimeUniversity,Shanghai201306,China;chen_lei1@163.com*Correspondence:tohuangtao@126.com(T.H.);cai_yud@126.com(Y.-D.C.);Tel.:+86-021-

4、6613-6132(Y.-D.C.)†Theseauthorscontributedequallytothiswork.Received:3January2018;Accepted:6March2018;Published:12March2018Abstract:Breastcancerisoneofthemostcommonmalignanciesinwomen.Patient-derivedtumorxenograft(PDX)modelisacutting-edgeapproachfordrugresearchonbreastcan

5、cer.However,PDXstillexhibitsdifferencesfromoriginalhumantumors,therebychallengingthemolecularunderstandingoftumorigenesis.Inparticular,geneexpressionchangesaftertissuesaretransplantedfromhumantomousemodel.Inthisstudy,weproposeanovelcomputationalmethodbyincorporatingsevera

6、lmachinelearningalgorithms,includingMonteCarlofeatureselection(MCFS),randomforest(RF),androughset-basedrulelearning,toidentifygeneswithsignificantexpressiondifferencesbetweenPDXandoriginalhumantumors.First,831breasttumors,including657PDXand174humantumors,werecollected.Base

7、donMCFSandRF,32geneswerethenidentifiedtobeinformativeforthepredictionofPDXandhumantumorsandcanbeusedtoconstructapredictionmodel.ThepredictionmodelexhibitsaMatthewscoefficientcorrelationvalueof0.777.Seveninterpretableinteractionswithintheinformativegeneweredetectedbasedonthe

8、roughset-basedrulelearning.Furthermore,theseveninterpretableinteractionscanbewellsupportedbyprev