Predicting Protein-protein Interaction And Study On Hub Protein Classification And Interaction Law

Post-genomic era, study of protein-protein interaction and its network has become an important task of system biology. Developing theoretical methods to predict protein-protein interaction is not only a useful supplement to experiment, but also for researching the mechanism of protein-protein interaction. And study on interaction network can be used to understand the rules and characteristics of the evolution and functions of life.The PPI depends on space structures of proteins. As a basic element of protein structure, secondary structures should contain information of PPIs. In this study, a new approach was introduced to predict interaction of proteins solely by analyzing their secondary structures, and the predicting model was built based on Bayesian classifier. The model achieved a significant performance with prediction accuracy of 59.16% and 59.85% in open test for S. cerevisiae, human and in independent set test for mouse and D. melanogaster respectively. The result show that the relationship between secondary structures of proteins and Bayesian classifier can be used to predict protein-protein interaction.Hub protein, as a kind of protein with high connected degree, plays an important role in biological processes. However,connect degree can not describe the protein role in biological network, since hub proteins with equal or similar connect degree are usually not equal important. Using X-means clustering, hub proteins were classified as three categories based on biological annotation information in GO. Results indicate that the distribution of sub-networks of system Hub and non-Hub proteins is uniform, that of process Hub and component Hub proteins are modular obviously. The parameter, Protein class interaction bias (PCIB), was introduced to describe the interaction bias between or within Hub proteins and non-hub proteins. Results shown that interaction bias among hub proteins, between non-hub proteins and hub proteins were strong, the values of PCIBs among non-hub proteins or between hub and non-hub proteins are very small.