A Protein-Protein Interaction Prediction Method Embracing Intra-protein Domain Cohesion Information

Recently, many computational methods for predicting protein-protein interaction (PPI) have been developed by utilizing domain-domain interaction or associated information. However, most of the methods lack of reflecting the collaboration effect of multiple do-mains to the prediction of PPI. In this paper, we develop a computational model that considers not only inter relationship between protein pair but also the intra-domain functional cohesion effect in PPI. In the computational model, a value assigning method to reflect the intra and inter collaboration devised and the computed values are stored in interaction significance (IS) matrix. Then an equation for PPI prediction is devised on IS matrix. For S. cerevisiae PPI data from DIP, MINT and IntAct, domain data from Pfam-A, the prediction method achieved 73.91% and 92.02% sensitivity and specificity respectively.