Abstract
We study methods of non-linear global optimization in the problem of molecular docking and present a new implementation of protein-ligand docking and preparation of molecules for docking in a new open source software Hess. It uses BFGS for local optimization and two different forms of Monte Carlo method for global optimization. We compare Hess to another open-source package, Smina, and find that the chosen methods give better results in both performance and binding energy (using the same Vinardo scoring function). Also, during preparation of some molecules Hess provides a more accurate computation of implicit hydrogens thanks to the dearomatization algorithm provided by the Indigo cheminformatics package.