Protein-Protein Docking Using a Brownian Dynamics Simulations Approach

Abstract

A successful protein-protein docking method provides theoretical understanding of how two or more proteins combine and interact with each other at the atomic level. It involves some unsolved problems partly due to the huge sampling space. Here we present an adapted Brownian Dynamics (BD) method used to predict the structure of protein complexes. The BD protein docking approach includes two steps, 1) global BD sampling; 2) local energy minimizations. In the first step, we run thousands of independent BD simulations to explore the entire possible conformational spaces of the protein complexes. The proteins are treated as two rigid bodies, and the translational and rotational motions are simulated for one of the proteins (protein II) around the other (protein I). The intermolecular forces and torques between proteins are given by the sum of electrostatic and exclusion forces. In the second step, we conduct local energy minimizations for all protein complexes obtained from the step one, and rank them by interaction energies. To reduce the computational costs for energy evaluations, we developed a grid-based force field to represent protein I and solvation effect. The rigid-body energy minimizations of the protein complexes are based on the downhill simplex method using the newly developed force field. The prediction quality of this newly developed BD protein docking approach is evaluated on a re-docking experiment for predicting the acetylcholinesterase-fasciculin complex (PDB entry 1FSS). The result shows that 100,000 independent BD runs generated 32797 protein complexes for the subsequent local energy minimizations. The root mean square deviation (RMSD) between the predicted lowest energy and the crystal structures is 0.17 A. In conclusion, this adapted BD protein docking approach could be used for prediction of other protein complexes, and help better understanding protein-protein interactions.