基于SVM和KNN的蛋白质耐热性分类.pdfVIP

228 and 2007．43(16) ComputerEngineeringApplications计算机工程与应用 基于SVM和KNN的蛋白质耐热性分类 丁彦蕊1t一，蔡字杰2．一，孙俊1，须文波1 DINGYan—ruil,2CAI Junl．XUWen—b01 Yu-jie2,3,SUN 1．江南大学信息工程学院，江苏无锡214122 2．江南大学工业生物技术教育部重点实验室．江苏无锡214036 3．江南大学生物工程学院，江苏无锡214036 1．SchoolofInformation 214122，China Technology，SouthernYangtzUniversity，Wuxi，Jiangsu ofIndustrial 214036，China 2．KeyLaboratory Biotechnology．Wuxi，Jiangsu 3．Schoolof 214036。China biotechnology．Wuxi，Jiangsu E-mail：yanrui_din976@yahoo．com．ca DING a1．Classificationof VectorMachinesand Yan-rui，CAIYu-jie，SUNJun，et thermostabilityusingSupport protein K-Nearest and Neighbors．ComputerEngineeringApplications，2007，43(16)：228-230 aminoacid as is classified VectoiMachines Abstract：Regardingcompositioneigenvector，proteinthermostability usingSupport result andK—Nearest isfoundthatthe of vectormachinesisbetterthanK-Nearest 10— Neighbors．It usingsupport Neighbors．The local are cal and are82．4％and the and 77．6％and accuracyglobalaccuracy 83．4％respectively．Butaccuracyglobalaccuracy K-Nearest oftwokindsofmethodscanboth thattheamino 79．9％respectivelyusing Neighbors．Thepredictionaccuracy prove acid isthemainfactorthatinfluencesthe composition proteinthermostability． acid Keywords：amino composition；SVM；KNN；proteinthermostability 摘 要：以氨基酸含量为特征向量，研究了SVM和KNN预测蛋白质耐热性的准确度。结果表明，基于SVM的分类效果较好，其局 预测率均表明氨基酸舍量是影响蛋白质耐热性的主要因素。 关键词