基于网络模块性的蛋白质序列聚类

Clustering Protein Sequences through Modularity Optimization

蛋白质的远同源性探测是结构基因组学和功能基因组学的主要研究任务之一。一些具有一定相似结构和功能、但序列相似性却较低的蛋白质组成蛋白质超家族,则远同源性探测问题等价于对蛋白质超家族的识别问题。作者提出了一种基于模块性的聚类算法ModuleFind,该方法通过最大化蛋白质网络的模块性来寻找具有较强集团结构的划分。在蛋白质结构分类数据库(SCOP)超家族层次上进行的实验表明,该方法得到的聚类结果更接近分类基准,且具有较高的F-测度值。

Remote homology detection between protein sequences is one of the principal research objectives in structural and functional genomics.Proteins with similar structure and function but low sequence similarity consist of protein superfamily.Therefore,the detection of remote homologues is the task of identifying protein superfamily.In this manuscript,a clustering algorithm,called ModuleFind,based on network modularity was presented.The method maximizes the modularity of protein network to find the partitioning with strong community structure.The resulting algorithm gives high quality of clusters quantified by F-measure that combines precise and recall,in the experiments of the detection of the remote homologues based on the superfamily level of SCOP database.