GPU-accelerated Exhaustive Search of Side-Chain Conformations: An Application to Protein Docking
Taras A. Dauzhenka, Petras J. Kundrotas, Ilya A. Vakser
Predicting 3D structures of protein-protein complexes (protein docking) is one of the central problems in computational structural biology and computer-aided drug design. Due to a very large number of possible conformations, accounting for conformational changes upon protein binding is a computationally challenging task. The General-Purpose Graphic Processing Units (GPUs) appeared on the market in recent years and have been of great use to computationally intensive scientific applications. In this study, we implemented a GPU-based protocol that enables exhaustive search through hundreds of millions of side-chain conformations in just a few hours on a single desktop (GPU Quadro K4000, CPU Intel Xeon 2.7GHz). We demonstrated applicability of this protocol to the protein docking refinement problem and compared the results to the existing modeling procedures.