基于拓扑势加权的动态PPI网络复合物挖掘方法  

Mining Protein Complexes Based on Topology Potential Weight in Dynamic Protein-Protein Interaction Networks

作　　者：雷秀娟[1] 高银 郭玲[2] 

机构地区：[1]陕西师范大学计算机科学学院,陕西西安710119 [2]陕西师范大学生命科学学院,陕西西安710119

出　　处：《电子学报》2018年第1期145-151,共7页Acta Electronica Sinica

基　　金：国家自然科学基金(No.61672334;No.61502290;No.61401263);陕西省工业科技攻关项目(No.2015GY016);中国博士后科学基金(No.2015M582606)

摘　　要：从动态蛋白质相互作用(PPI)网络中挖掘蛋白质复合物是当前复合物挖掘研究的一个热点,但是目前大都采用未加权网络进行聚类分析,由于不能准确地描述网络的拓扑特性,因此其正确率不高.鉴于此,本文提出采用拓扑势场的方法来构造加权网络,网络中的每一个蛋白质都被视作一个物理粒子,在它周围存在一个虚拟的作用场,由此网络中所有蛋白质的相互作用联合形成一个拓扑势场,文中定义了结点间的拓扑势的概念,并以此来构造加权网络,之后采用马尔科夫聚类算法在DIP数据和Krogan数据上进行复合物挖掘.与其它经典算法相比,该方法的precision和f-measure值较高,能更好地识别蛋白质复合物.At present,many researchers focus on identifying protein complexes in the dynamic protein-protein interaction（ PPI） network. But most of the methods have been used in unweighted network, the accuracy is not high since they could not accurately describe the network topology characteristics. In this paper,we put forward the topology potential to construct the weighted network. A protein can be regarded as a physical particle in the network which has a virtual field around it, and the interaction of all proteins forms a topology potential field. By computing the value of topology potential between nodes, the weighted network is constructed, and Markov clustering algorithm is used to identify protein complexes. The experimental results compared with the classic algorithms on DIP data and Krogan data indicate that their precision and fmeasure value are higher and the proposed algorithm is more suitable to identify the protein complexes.

关 键 词：蛋白质相互作用网络 拓扑势 马尔科夫聚类 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]