ParaDockS: a framework for molecular docking with population-based metaheuristics.

Molecular docking is a simulation technique that aims to predict the binding pose between a ligand and a receptor. The resulting multidimensional continuous optimization problem is practically unsolvable in an exact way. One possible approach is the combination of an optimization algorithm and an objective function that describes the interaction. The software ParaDockS is designed to hold different optimization algorithms and objective functions. At the current stage, an adapted particle-swarm optimizer (PSO) is implemented. Available objective functions are (i) the empirical objective function p-Score and (ii) an adapted version of the knowledge-based potential PMF04. We tested the docking accuracy in terms of reproducing known crystal structures from the PDBbind core set. For 73% of the test instances the native binding mode was found with an rmsd below 2 A. The virtual screening efficiency was tested with a subset of 13 targets and the respective ligands and decoys from the directory of useful decoys (DUD). ParaDockS with PMF04 shows a superior early enrichment. The here presented approach can be employed for molecular docking experiments and virtual screenings of large compound libraries in academia as well as in industrial research and development. The performance in terms of accuracy and enrichment is close to the results of commercial software solutions.