基于模块度优化的蛋白质网络集团探测与分析

Detecting and analyzing of communities in protein-protein interaction network based on modularity

探测蛋白质相互作用网络中的功能模块对于理解生物系统的组织和功能具有重要的意义。目前,普遍的做法是将蛋白质相互作用网络表示成一个图,利用各种图聚类算法来挖掘功能模块。本文采用了基于模块度优化的图聚类算法来探测蛋白质相互作用网络中的集团,从具有2617个节点11855个相互作用的酵母蛋白相互作用网络中探测出68个集团。对于得到的集团,首先从拓扑结构的角度验证其的确是内部连接稠密的子图,然后分析了MIPS数据库中ComplexCat提供的已知的蛋白质复合体与这些集团的重叠情况,发现很多蛋白质复合体完全包含在某些集团中,最后使用超几何聚集分布的P值来分析一个集团对某个特定功能的富集程度,并根据最小的P值对应的功能来注释该集团的主要功能,发现集团中大部分的蛋白质具有相同的功能。研究结果表明,该方法探测的集团具有重要的生物学功能意义。

Detecting functional modules in protein-protein interaction （PPI） networks is very important to understand the organization and function of the biological system. At present, a common method of revealing functional modules in PPI networks is graph clustering where PPI networks are modeled as a graph in which vertices represent proteins and edges represent interactions. Here, a modularity-based method was used to find communities in PPI networks. Using this method, 68 communities were detected from a network involving 11 855 interactions among 2 617 proteins in yeast. For the communities outputted by the method, we firstly assessed the validity from the topology perspective and found that they are densely connected local subgraphs. Then, we matched known protein complexes annotated by ComplexCat database in MIPS against these communities and found that known protein complexes are largely contained in them in their entirety. At last, we used hypergeometric distribution P-value to measure whether a community is enriched with proteins from a particular category more than would be expected by chance. We assigned each community the main function with the lowest P-value in all categories and found that most proteins in the same community have the same function. Tests show that communities revealed are with significant biological functions.