Research On Algorithm Of Identifying Protein Complexes Based On Protein-protein Interaction Network

Protein complex which plays an important role in the process of biological function is one of the most basic units in biological cellular.So researching the process of forming complex protein based on protein interactions as well as performing specific biological function in the form of protein complex is a great challenge for the biological researchers.Meanwhile,identifying protein complexes from protein-protein interaction network is of vital significance for explaining the rule of biological function.A method for reconstruction protein-protein interaction network relied on connectivity theory is proposed to optimize the high false positive rate and false negative rate exist in protein-protein interaction network.This method calculates interaction weights by using the eigenvalues and eigenvectors of the adjacency matrix of the network.Whether an interaction exists in reconstruction network depends on its interaction weight.We compare four efficient algorithms operating in initial protein-protein interaction network and in new network reconstructed based on connectivity theory.The comparison results show that the protein complexes predicted from graph connectivity network obviously outperform that in initial network which implies that connectivity theory effectively enhances the reliability of the interactions in protein-protein interaction network.This paper introduces a recognition algorithm for protein complexes,called CFPE,based on edge clustering coefficient under the framework of cell-core-attachment.The proposed algorithm first searches the cell structures in protein-protein interaction network.Then cells are expanded to the complex cores as the heart of complexes.And lastly,we add attachments into the complex cores for generating whole complexes.The experiment results show that cell is rich for both topological property and biological significance.What is more,CFPE method has higher prediction F-measure and proteins coverage rate than the comparison algorithms.And the percentage of predicted protein complexes of marked biological significance is also superior to other comparison algorithms.