Combining local graph clustering and similarity measure for complex detection

Protein complexes are key modules to perform protein functions within protein-protein interaction (PPI) network. Protein complexes are determined by both topological and biological properties. The information from protein primary sequence can help to understand principles of cellular organization and function of complexes. In this paper, a novel method for detecting protein complexes from protein amino acid sequence has been presented. A simple feature representation from protein primary sequence is presented and become a novel part of feature extraction. In searching process, similarity measure is applied to detect protein complexes as one of the constraints. First, the comparison between our method and other three competing methods is performed on the two different Yeast PPI networks. Second, we validate the detected complexes using function analysis. The experimental results show that our method outperforms other three methods on the number of detecting real complexes. In addition, our method can provide an insight into the further biological study.