Development of a Protein-Ligand Docking Program Based on Global Optimization

Abstract

Protein-ligand docking has become an essential tool for computer-aided drug discovery since docking programs were first developed in 1980s. The goals of docking are to predict 1) the binding mode and 2) the binding affinity of a given protein-ligand complex accurately. Accurate prediction of binding mode requires appropriate sampling of both protein and ligand conformations. Many available docking programs sample ligand structures successfully because ligand has a relatively small number of degrees of freedom. However, a lot of current docking programs treat receptor as a rigid molecule although receptor often adapts its shape to bound ligand because treating receptor flexibility is a very complicated problem. First of all, the large conformational space of receptor is a challenge for typical sampling methods. In addition, current energy functions such as empirical docking score functions or force field-based energy functions do not accurately describe flexible receptor-flexible ligand interactions yet. In this thesis, the development process of an efficient docking program that treats receptor flexible, called GalaxyDock, is described. A powerful global optimization technique, called conformational space annealing, was employed for simultaneous sampling of the conformational space of protein and ligand. In addition, a new energy function for flexible-receptor docking was designed by combining the AutoDock energy function and a knowledge-based ROTA potential. With these components for sampling and scoring, GalaxyDock shows high performances in the binding pose prediction and virtual screening benchmark tests when compared to other state-of-art docking programs. This result suggests that the GalaxyDock program can provide a firm basis for further method developments and for practical applications to in sillico drug discovery processes.