An Enhanced Evolutionary Algorithm for Detecting Complexes in Protein Interaction Networks with Heuristic Biological Operator

Abstract

Detecting complexes in protein interaction networks is one of the most important topics of current computational biology research due to its prominent role in predicting functions of yet uncharacterized proteins and in diseases diagnosis. Evolutionary Algorithms (EAs) have been adopted recently to identify significant protein complexes. Conductance, expansion, normalized cut, modularity, and internal density are some well-known examples of complex detection models. In spite of the improvements and the robustness of predictive functions introduced by complex detection models based on EA and regardless of the general topological properties of protein interaction networks, inherent biological data of protein complexes has not, or rarely exploited and incorporated inside the methods as a specific heuristic operator. The aim of this operator is to guide the search process towards discovering hyper-connected and biologically related complexes by allowing a more effective exploration of the state space of possible solutions. Thus, the main contribution of this study is to develop a heuristic biological operator based on Gene Ontology (GO) annotations where it can serve as a local-common optimization approach. In the experiments, the performance of eight EA-based complex detection models has analyzed when applied on the yeast protein networks that are publicly available. The results give a clear argument for the positive effect of the proposed heuristic biological operator to considerably enhance the reliability of the current state-of-the-art optimization models.