Effect of preprocessing and simulation parameters on the performance of molecular docking studies

Context Molecular docking is an important and rapid tool that provides a comprehensive view of different molecular mechanisms. It is often used to verify the binding interactions of many pairs of molecules and is much faster than more rigorous approaches. However, its application requires carefully preprocessing each molecule and selecting a series of simulation parameters, which is not always done correctly. We show how preprocessing and simulation parameters can positively or negatively impact molecular docking performance. For example, the inclusion of hydrogen atoms leads to better redocking scores, but molecular dynamics simulations must be performed under certain constraints; otherwise, it may worsen performance rather than improve it. This study clarifies the importance and influence of these different parameters in the simulation results. Methods We analyzed the influence of different parameters on the predictive ability of molecular docking techniques using two software packages: AutoDock Vina and AutoDock-GPU. Thus, 90 receptor-ligand complexes were redocked, evaluating the root mean square deviation (RMSD) between the original position of the ligand (receptor-ligand complex obtained experimentally) and that obtained by the software for every analysis. We investigated the influence of hydrogen atoms (on the receptor and on the receptor-ligand complex), partial charges (QEq, QTPIE, EEM, EEM2015ha, MMFF94, Gasteiger-Marsili, and no charge), search boxes (size and exhaustiveness), ligand characteristics (size and number of torsions), and the use of molecular dynamics (of the receptor or the receptor-ligand complex) before docking analyses.