Acceleration of protein-ligand docking simulation using graphics processing units

概要

論文の詳細を見る
In modern drug development, it is important to obtain ligand (small molecule) which binds specifically to a target protein related to disease. A common structure-based screening approach is to use molecular docking simulation to identify the docking between a target protein and ligand. This approach must simulate a docking to predict whether a target protein binds to ligand for tens of millions compounds. Screening a library of compounds is enormously costly and time-consuming. When three-dimensional (3-D) structure of the target protein is already known, molecular docking simulation between a ligand and a protein can be performed in order to search large ligand libraries for drug-candidates: ligands that bind to this protein with high affinity. In this paper, we show how AutoDock, which is software for simulating the docking between a target protein and ligand using a genetic algorithm, can calculate the genetic algorithms scoring function in parallel. The methods presented for parallelizing the workload result in an average speedup of 3.3 times on a comparative CUDA enabled graphics processing unit (GPU)(CUDA: Compute Unified Device Architecture).
2013-07-15