Research On Identification Of Biological Function Modules In Tissue-Specific Protein-Protein Interaction Network

Identification of key proteins and functional modules in the Tissue-Specific Protein-Protein Interaction(PPI) networks has an important role on understanding the intrinsic mechanisms of biological activities,targeted disease diagnosis and drug design. Identification of proteins and modules require high reliability of PPI network,recognition algorithm with high accuracy,and systematic biological analysis tools. In the paper,we used the human tissue-specific genes and their corresponding PPI networks as initial data source,identified important Date Hub(DH) proteins,and verified biological functions of these proteins and their associated intensive modules. And,we further researched sub-graph query algorithms among multiple PPI networks for exploring the conserved sub-graph. Finally,we proposed a Semi-Markov Random Walk based on iterative weighted algorithm for sub-graph query. The contributions of the paper include the followings:Firstly,for existing human tissue-specific genes and PPI networks,we used the shortest path algorithm to design a construction algorithm of tissue-specific PPI network which achieved by MapReduce framework. We made experimental verification for PPI networks of human kidney and brain,and provided a quantitative evaluation of the reliability of the network.Secondly,for analysis of tissue-specific functions and key proteins found in PPI networks,we designed the algorithm of identifying functional modules related to DH proteins on the basis of merging topological distance and GO similarity. From the PPI network of human kidney,we identified 118 DH proteins. Then we selected 36 strong association DH proteins to make analysis. Experiments showed that the bridge role of the IGF-1 and its strong relation modules in renal hypertrophy process.Finally,in order to explore the conserved modules among different human tissues,we proposed a weighted sub-graph search algorithm based on semi-Markov random walk model and iterative reduction. Calculating node similarities and reducing sizes of the target graph are two common means for improving query precisions and reducing computational complexity. This paper presents a sub-graph query algorithm based on semi-Markov random walk model to query sub-graph in complex protein interaction networks. A comprehensive similarity measurement based on semi-Markov random walk model is designed to integrate the similarities of nodes,structures and their neighbors. Meanwhile,an iterative procedure is applied to reduce the size of targeted graph by removing nodes with lower similarities. The experimental results on multiple real protein query networks demonstrate that the proposed algorithm improves its performance both in query precisions and computational complexity.