基于加权网络和局部适应度的蛋白质复合物识别算法

蛋白质复合物识别对分析蛋白质网络的结构特征和模块功能具有重要意义。通常在蛋白质网络中挖掘稠密子图或模块来识别其中的蛋白质复合物,限制了其应用范围和识别的准确性。针对该问题,提出了一种基于加权网络和局部适应度的蛋白质复合物识别算法,该算法综合稠密子图的密度指标和模块性定义了新的局部适应度函数,并基于边聚集系数构建加权的蛋白质网络,根据权值选择边,在加权蛋白质网络中将种子边不断聚类扩展,从而获取具有最大综合适应度的子图作为蛋白质复合物。在酵母蛋白质等多个实际网络中试验表明,该算法能够有效提升蛋白质复合物识别的准确性。

An algorithm for identifying protein complexes based on weighted network and local fitness

It is of great significance to identify protein complexes for understanding the structure and module function f protein networks. Usually, the ways about mining dense subgraph or module would limit their scope of application and recognition accuracy on protein complexes identification. To solve this problem, this paper proposes a novel protein complex recognition algorithm based on weighted network and local fitness. Integrating density of subgraph and modularity, we define a new local fitness function, and use edge clustering coefficient to construct the weighted protein network, select some seed edges according to their weights, then extending the clustering around the seed edge until gaining a biggest protein subgraph with maximum comprehensive fitness. Experiments in yeast protein networks show that, our algorithm can effectively improve the accuracy on protein complexes identification.