用拓扑和语义信息计算熵值的PPI网络聚类  

Clustering protein-protein interactions network by computing entropy with topological and semantic information

作　　者：徐雷[1] 钟诚[1] 李智[1,2] 

机构地区：[1]广西大学计算机与电子信息学院,广西南宁530004 [2]广西科技信息网络中心,广西南宁530022

出　　处：《计算机工程与设计》2016年第3期757-761,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(61462005);广西自然科学基金项目(2014GXNSFAA118396);广西研究生教育创新计划基金项目(YCSZ2014034);广西教育厅-广西大学博士点建设基金项目(P11900119)

摘　　要：采用较大相关种子簇方法优化种子选择,获得更佳的起始聚类,在种子生长过程中忽略蛋白质交互网络(PPI网络)中关联较小的边,筛选掉基因术语相似度较低关系连接,提升算法的效率与聚类结果的精确度,通过融合拓扑结构和语义相似度优化计算熵值,提高熵值代表的信息量,获得更佳的PPI网络聚类结果。实验结果表明,与已有算法相比,该方法能获得更大f-score的聚类结果,聚类时间更短。The larger cluster of relative seeds was used to optimize the seed selection to obtain better initial clustering,the edges with small correlation part in seed growth process were ignored and the links with lower similarity between gene ontology terms were filtered in protein-protein interaction networks（PPI networks）to improve the efficiency and precision of the clustering results.The amount of information in entropy was enhanced and better clustering results for PPI networks were obtained using the topological structure and semantic similarity to optimize the computation of entropy.The experimental results show that,compared with the existing algorithms,the proposed method is more efficient and achieves better clustering results with bigger value of f-score.

关 键 词：蛋白质交互网络 聚类 熵 语义相似度 拓扑结构 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]