基于K近邻的蛋白质功能的预测方法  

Prediction of Protein Functions based on K Nearest Neighbors Method

作　　者：倪青山[1] 王正志[1] 黎刚果[1] 孟祥林[1] 

机构地区：[1]国防科技大学机电工程与自动化学院,湖南长沙410073

出　　处：《生物医学工程研究》2009年第2期87-90,共4页Journal Of Biomedical Engineering Research

基　　金：国家自然科学基金项目(60471003)

摘　　要：蛋白质功能预测是后基因组时代研究的重要问题之一。利用蛋白质相互作用网络,提出了一种基于K近邻的蛋白质功能的注释方法,该方法首先计算待注释的蛋白质与所有已知功能的蛋白质间的注释环境相似度,选择其中最相似的K个蛋白质,将该K个蛋白质的功能注释进行加权平均,作为待注释的蛋白质最终的功能注释。在构建的芽殖酵母的两组大规模相互作用数据集上的测试表明,该方法能够有效的对蛋白质功能进行预测,在蛋白质功能预测性能上优于现有的一些方法。In the post - genomic era, protein function annotation is one of the most important problems. In this paper, a new method, based on K Nearest Neighbors method, was proposed to predict protein functions from proteinprotein interactions. The similarity of protein annotation environments between the unannotated protein and all the annotated proteins were computed and the K proteins, which were most similar to the unannotated protein, were selected, then a weighted average of the functions of the K proteins was used as the functions estimate for the unannotated protein. The results, which is obtained by testing on the two large scale data sets constructed for Saccharomyces, show that the proposed method can predict protein functions effectively and outperforms some existing approaches.

关 键 词：蛋白质相互作用 蛋白质功能预测 K近邻算法 相似性 加权方法 

分 类 号：Q811[生物学—生物工程] TP391[自动化与计算机技术—计算机应用技术]