Identification of functional modules in dynamic weighted PPI networks by a novel clustering algorithm

Abstract

The Density-Based Spatial Clustering of Application with Noise algorithm (DBSCAN) suffers the limitations of selecting global parameters and having the low accuracy in recognizing overlapping protein complexes. In order to overcome the disadvantage of slow convergence and being vulnerable to trap in local optima in Artificial Bee Colony algorithm (ABC), we designed a method with novel weights and distance calculated which is suitable for network topology and the interaction between proteins. Furthermore, a truncation-championship selection mechanism (TCSM) was proposed to avoid local optimum when onlooker bees search nectar source. Meanwhile, we present the adaptive step strategy (ASS) to improve the clustering speed in ABC algorithm. Finally, in order to overcome the shortcoming which is unable to identify protein complexes in the DBSCAN algorithm, a strategy is proposed to optimize the clustering result. The experimental results on superior precision and recall parameters demonstrate that our method has competitive performance for identifying protein complexes.