Effect of ligand torsion number on the AutoDock mediated prediction of protein-ligand binding affinity

Abstract

Molecular docking simulation is a useful tool in the prediction of protein-ligand binding affinity on a large scale and has great potential in various application fields such as virtual screening of potential drug molecules. However, the reliability of molecular docking is still weak in the estimation of ligand-binding free energy, which limits the applicability of molecular docking simulation. Ligand torsion number is related to the flexibility of ligand and generally incorporated as a crucial variable in the thermodynamic function of binding free energy. In this study, we investigated how the ligand torsion number has influence on the binding affinity prediction of AutoDock, a popular molecular docking simulation tool. The pKd values of various protein-ligands were estimated by using the binding free energy function of AutoDock and compared with their experimental pKd values. The torsion number dependent comparison showed that the predicted binding affinities were mostly underestimated in the complexes of higher torsion numbers, whereas the underestimated and overestimated cases were relatively balanced at relatively lower torsion numbers. A new weight factor for torsion-free energy term of binding energy function was determined and introduced to make correction to the underestimation of binding affinity of ligands with high torsion numbers. It is expected that the torsion number dependent deviation pattern of AutoDock and its correction strategy are useful in the large-scale validation of protein-ligand binding affinity.