Research On Analysis Algorithm Based On Protein Interaction Networks

Protein is an important part of the organism,and it is also the material basis of life.It participates and controls most of the life activities in the organism.Proteins do not exist alone in living organisms,but they realize the specific functions of living organisms through the interactions between proteins.With the development of high-throughput experimental methods,a large number of protein-protein interaction(PPI)data has been accumulated,which provides a possibility for systematic research on the interactions between proteins,important functional modules detecting,and essential proteins identifying in the networks.The studies of functional modules and essential proteins on the basis of PPI networks can not only promote the research of life science,but also have important application in the diagnosis and treatment of disease and the design of drug target cells.In this paper,we study the analysis algorithms of PPI based on the network hierarchy structure.We mainly focus on the problem of mining the functional modules and predicting essential proteins by integrating multi-source biological information.The specific research includes the following three aspects:(1)A functional modules detecting algorithm FM-HS based on hierarchical structure partitioning in PPI network is proposed.The method uses a genetic algorithm to find a hierarchical tree with maximum likelihood corresponding to the PPI networks.Through the hierarchical division of the tree,the functional modules can be divided,and then the best partitioning scheme according to the modularity can be obtained.Furthermore,the possibility of protein-protein interactions can be obtained through the information of common ancestors in the tree between the nodes.The method can simultaneously mine functional modules and predict protein-protein interactions.Experimental results on standard database show that the proposed FM-HS algorithm can more accurately mine the functional modules in the PPI networks.(2)An essential protein identifying algorithm(EPM)based on Markov random walk is proposed.The method considers both the biological information and the network topology information and can overcome the negative impact of high data noise.The method EPM uses the Markov random walk to assign a score representing its importance to each vertex in the PPI network,and the scores of all vertices constitute the n column vector.We initialize the score,and then the score can be randomly walked in the network according to a certain probability and modified in the transmission.Finally,after sorting the scores in descending order,the output scores corresponding k proteins are the essential proteins.The experimental results on the standard dataset show that the proposed EPM algorithm can identify major essential proteins more accurately.(3)An algorithm for identifying essential proteins based on genetic algorithm in PPI networks is proposed.The algorithm selects m initial individuals,and each individual is composed of P proteins.The genetic algorithm is used to measure the critical of the top-P proteins.The individual with the most essentiality(ie,the maximum fitness function value)is selected.Finally,the local optimization is carried out on the individual solution.In the fitness function,multi-source biological information such as gene expression data and domain interaction strength is integrated into this method,and the influence of protein second-order neighbors on the essentiality of the protein is taken into account.Experimental results on standard datasets also show that compared with other classical algorithms,the EPGA algorithm we proposed has higher accuracy in identifying essential proteins.