A novel molecular docking program based on a multi-swarm competitive algorithm

Molecular docking plays a vital role in the pharmaceutical discovery field. As a structure-based drug design tool, it predicts the lowest energy conformation for ligands (small molecules) binding in the active region of receptors (macromolecules) based on their three-dimensional crystal structures. The satisfactory accuracy and efficiency of docking programs depend on two components: a robust sampling algorithm and a precise scoring function. The former searches for candidate conformations of ligands and receptors, and the latter evaluates and ranks these conformations. Various docking programs have been proposed, employing different sampling algorithms and scoring functions. In this study, a novel molecular docking program is proposed, which uses a multi-swarm competitive algorithm as the search method to improve the performance of the docking problems, especially with highly flexible ligands. The algorithm adopts a multi-swarm framework and feedback mechanism to realize the competition between the swarms, and incorporates a gradient descent algorithm to refine the current population. The CASF-2016 benchmark dataset with 285 complexes is used to verify the proposed docking programs, and the experimental results reflect that it can provide very competitive docking performances when compared with other commonly used programs.