基于支持向量机的蛋白质相互作用识别  

Identifying Protein-Protein Interactions Based on SVM

作　　者：陆林英[1] 魏雅卓[1] 崔颖[1] 孙平平[1] 马雅楠[1] 马志强[2,1] 

机构地区：[1]东北师范大学计算机学院,长春130117 [2]吉林大学计算机学院,长春130000

出　　处：《生物信息学》2009年第4期255-257,共3页Chinese Journal of Bioinformatics

基　　金：教育部应用统计重点实验室资助

摘　　要：支持向量机(SVM)是广泛应用于各个领域的分类算法,包括生物信息学。本研究应用SVM作为蛋白质相互作用的分类算法,所用蛋白质相互作用数据下载于墨尼黑生物信息学中心的酿酒酵母数据集,包含有6 736条蛋白质,其中相互作用的有4 837对,不相互作用的有9 674对。提取蛋白质主要结构的电荷和等电位点特征,并应用SVM分类算法对此进行了分类。结果显示,分类的正确率在60%左右,但是较系统发育谱法还是获得了较高的分类正确率。SVM classification algorithm has been applied to identify protein - protein interactions data. SVM has been employed widely in many fields, such as bioinformatics. The protein interactions data download from munich information center for protein sequence, which includes 6736 proteins in Scerevisiae yeast dataset. Interaction protein has 4837 pairs and non - interaction protein has 9674 pairs. Extracting charge and the isoeleetrie point （PI） as mainly features of protein primary structure, and SVM has been used to classify interaction proteins.Although the accuracy of classification was about 60%, comparing with Phylogenetic profiling, it also got good results.

关 键 词：蛋白质相互作用 SVM 序列主要结构 

分 类 号：Q811[生物学—生物工程]