Probabilistic models for capturing more physicochemical properties on protein-protein interface

Protein-protein interactions play a key role in a multitude of biological processes, such as signal transduction, de novo drug design, immune responses, and enzymatic activities. It is of great interest to understand how proteins interact with each other. The general approach is to explore all possible poses and identify near-native ones with the energy function. The key issue here is to design an effective energy function, based on various physicochemical properties. In this paper, we first identify two new features, the coupled dihedral angles on the interfaces and the geometrical information on π-π interactions. We study these two features through statistical methods: a mixture of bivariate von Mises distributions is used to model the correlation of the coupled dihedral angles, while a mixture of bivariate normal distributions is used to model the orientation of the aromatic rings on π-π interactions. Using 6438 complexes, we parametrize the joint distribution of each new feature. Then, we propose a novel method to construct the energy function for protein-protein interface prediction, which includes the new features as well as the existing energy items such as dDFIRE energy, side-chain energy, atom contact energy, and amino acid energy. Experiments show that our method outperforms the state-of-the-art methods, ZRANK and ClusPro. We use the CAPRI evaluation criteria, I_rmsd value, and F_nat value. On Benchmark v4.0, our method has an average I_rmsd value of 3.39 Å and F_nat value of 62%, which improves upon the average I_rmsd value of 3.89 Å and F_nat value of 49% for ZRANK, and the average I_rmsd value of 3.99 Å and F _nat value of 46% for ClusPro. On the CAPRI targets, our method has an average I_rmsd value of 3.56 Å and F_nat value of 42%, which improves upon the average I_rmsd value of 4.27 Å and F_nat value of 39% for ZRANK, the average I_rmsd value of 5.15 Å and F_nat value of 30% for ClusPro. © 2014 American Chemical Society.