Relative Contribution of the Refinement Steps to the Protein-Protein Docking Success Rate

Abstract

Structural modeling of protein-protein complexes is a challenging task as proteins may undergo significant conformational changes upon binding. Flexible docking of proteins typically consists of low-resolution rigid-body global docking search, followed by the structural refinement that takes into account conformational flexibility. The refinement typically involves consecutive steps associated with different weakly coupled sets of degrees of freedom (DOF). Previously, we developed a docking refinement pipeline consisting of three steps: (i) rigid-body adjustment of the protein position, which minimizes a semi-empirical atom-atom contact potential (rigid-body rotational/translational DOF), (ii) conformer selection procedure applied to the backbone loops at the interface, to maximize the Ramachandran probability (backbone loops torsional DOF), and (iii) repacking of the interface side chains using a library of rotamers, to minimize the steric clashes (side-chain torsional DOF). In this study, we investigate relative contributions of these steps to the refinement success, in terms of the fraction of the native contacts, ligand RMSD, and interface RMSD, for bound and unbound proteins in the Dockground X-ray docking benchmark set 4.0 (http://dockground.compbio.ku.edu).