Research On Protein Complex Mining Algorithm In PPI Network

Mining protein complexes from protein-protein interaction networks is one of the significant researches in the proteomic.It helps people understand the characteristics of organism and analysis disease of mechanism better by studying complexes and function modules.However,the available data has high dimension,false positives and false negatives,and the protein-protein interaction is not stable,which have affected the accuracy of detecting complexes,and it is too single to determine the relationship of nodes in terms of one feature,different features affects the final prediction results a great extent.This article aims to mine protein complexes in PPI network,the main studies are as follows:(1)Identifying complexes based on cuckoo search mechanism.This paper considers the topological characteristics,biological characteristics and relationship of node and complexes to evaluate the similarity between nodes,then clustering for twice,which can avoid clustering by single characteristic of node.The first stage,cuckoos are more willing to seek higher similarity of nest to brood,which makes it has more survival chance and uneasily found by host.In this paper,it calculates the similarity of bird and the nest firstly,the birds try to find the ideal nest so that we get the initial cluster through multiple iterations.Then the birds,who have not found nest,have been clustered according to the affinity density.The second stage,clustering the complexes,which have less nodes or the nodes have not gathered yet.Merging the node and its neighbors whose cluster density is greater than the threshold value.Then expanding them with affinity density,which makes the sparse nodes cluster as possible and obtain more complexes.This method considers different measurements to mine complexes and it is more effective in clustering.(2)This paper uses the thought of granular computing to detect protein complexes.Based on the natural commonalities of granular computing and clustering,the theory of quotient space is introduced to mine complexes.Firstly,problems expressed by subset or quotient space and different quotient space is embodied the quotient set of different granular.Then choosing appropriate space to gather based on different divisions through equivalence relation and composing the different granularity space.Finally,it can get the solution of original problems.In this paper,the method estimates the relationship between particles to make up for the inadequacy of data in combination with the PPI data and Gene Ontology data,and dealing with the network based on theory of quotient space.Granulating the network to construct the quotient space and graining the ultima results through merging layer by layer.Then also saving the particles un-clustered in the class each time to increase the diversity of particles,the final result is obtained after purification.It turned out that,the proposed method could exploit protein complexes more accurately and efficiently.(3)This paper uses the topological potential weighted to mine protein complexes in dynamic PPI networks.PPI data could not accurately describe the real network because of the false positives and false negatives.The proposed method adopts the topological potential to optimize the network.It has the similar properties with nodes of PPI network that each particle interacts with others in topology potential field.Thus,it can better reflect the biological structure of the PPI network that we construct the weighted network through topological potential between nodes,at the same time,combined with gene expression data to construct sequential sub-network.Then,Markov clustering algorithm has been applied to identify protein complexes on PPI data.The experimental results show that our method is more suitable to identify the protein complexes compared with other classic algorithms.