PROTEIN-PROTEIN DOCKING REFINEMENT USING RESTRAINT MOLECULAR DYNAMICS SIMULATIONS

Abstract

A realistic prediction of the structure of protein-protein complexes is of major importance to obtain three-dimensional models for the interaction of proteins to form complexes and assemblies. In addition to the systematic search for putative binding sites on the surface of two binding partners, the second step of a docking effort, the refinement of docked complexes, is a major bottleneck to obtain realistic interaction geometries. Typically, the first initial systematic search employs rigid partner structures or few flexible degrees of freedom, whereas the refinement step involves fully flexible partner structures. The possibility to refine docked complexes using restraint MD simulations combined with an implicit solvent (Generalized Born) model was explored on three example test complexes starting from unbound partner structures. Significant improvement, both in scoring and agreement with the native complex structure after refinement was observed for two test cases. No improvement was found for a test case of a complex with lower binding affinity. The method can be easily applied to any docked protein-protein complex, however, more general applicability requires further improvements in the scoring function.