Study On Law And Predicting Methods Of Protein-protein Interaction

As the main way of function execution of proteins, Protein–protein interactions (PPIs) play crucial roles in a variety of biological processes. One of the most important tasks in resent biology researches is to explore the mechanism of PPI and distinguish whether particular proteins could interact with each other. It is of practical significance to study on the law of PPI based on the sub-cellular and structural information of proteins, as well as to construct prediction model.By defining a parameter describing Protein-protein Interaction Relative Bias (PIRB), the law of protein interaction between or within organelles, membranes of yeast cell and different classes for S. cerevisiae, C. elegans and E. coli was investigated based on the database of interacting protein (DIP), Gene Ontology (GO), Structural classification of proteins (SCOP) and related databases such as SWISS-PROT. The results indicate that there exists obvious bias of protein-protein interaction between or within different organelles, membranes and protein's structural classes. The biological implication and the law of PPI were discussed in this work.There is strong relationship between PPI and the structure information of protein. In this work, secondary structure data is used as feature in the prediction of PPI and protein subunits interaction. In the work of subunits binding interaction prediction, the information of protein's secondary and super-secondary structure was analyzed and used to construct predicting models by using support vector machine (SVM). This model achieved a significant performance with total accuracy of 64.52%, sensitivity of 66.87% and correlation coefficient of 0.291.In the work of PPI prediction, both the structural data and its related regional location are considered in model construction. The predicting model which is constructed by using SVM as well gained a high accuracy of 88.01% and the correlation coefficient is 0.761 when it was applied to S. cerevisiae. The prediction software and all data sets used in PPI prediction are freely available at http://ibt.imust.cn/PPIP.html.