异源蛋白质相互作用数据整合算法的进展  

Advances in algorithms applied on various protein-protein interaction data sources integration

作　　者：王文馨[1] 陈宇光[1] 石铁流[2] 

机构地区：[1]上海大学生命科学学院,上海200444 [2]中国科学院上海生命科学信息中心,上海200031

出　　处：《生命科学》2008年第5期821-826,共6页Chinese Bulletin of Life Sciences

基　　金：“973”项目(2007CB108800);“863”项目(2006AA02Z313);国家自然科学基金(90408010)

摘　　要：蛋白质相互作用在生物学过程和细胞功能行使中起核心作用。高通量技术的应用结合计算机预测方法的发展,使得直接和间接来源的蛋白质相互作用数据得到了大规模的增加。如何系统地整合这些数据并从中提取有用的信息是一项挑战,这也促使了许多整合算法应运而生。本文综述了八种整合蛋白质相互作用数据源的方法:投票、支持向量机、朴素贝叶斯、逻辑斯蒂回归、决策树、随机森林、基于随机森林的k-近邻法以及混合属性分类等方法。Protein-protein interactions are crucial for all biological processes and fundamental to virtually every aspect of cellular functions. Developments of high through-put experimental techniques and in silico prediction methods help to increase direct and indirect protein-protein interactions data. How to systematically integrate those data and extract the meaningful information from them is a really challenge. Many computational ap- proaches are therefore emerging for the purposes. This review presents recent advances for the application of those approaches in integrating protein-protein interaction data sources, including voting, support vector machine, naive bayes, logistic regression, decision tree, random forest （RF）, RF-based k-nearest-neighbor and mixture of feature experts.

关 键 词：蛋白质相互作用 数据整合 二分类器 

分 类 号：Q51[生物学—生物化学] Q811.4