Accurate Protein-Ligand Binding Free Energy Estimation by Applying QM/MM Calculations on Multi-Conformers Predicted from a Classical Mining Minima Approach

Abstract In a bid to develop a more accurate binding free energy prediction method that is becoming ever more important in drug discovery, we have concocted four protocols that combine QM/MM calculations and the mining minima method. We tested these protocols on 9 targets and 203 ligands. Our protocols carry out free energy processing with or without conformational search on the selected conformers obtained from classical VM2 calculations, where their force field atomic charge parameters are substituted with those obtained from a QM/MM calculation. The free energy processing with or without conformational search takes an average of 15 minutes or 1 minute per conformer, respectively. A reasonably high Pearson’s correlation coefficient (0.69) was achieved in relation to experimental binding free energies across a broad range of targets, demonstrating the generality of the method. Our proposed differential evolution algorithm, with a universal scaling factor of 0.2, achieved a low mean absolute error (MAE) of 0.60. This correlation and MAE are noticeably better than many common methods and comparable with most popular relative binding free energy (RBFE) methods. Meanwhile, the computing cost of our protocol is significantly lower than any of them. Moreover, our method gives binding poses of local minima, which can be further exploited for protein-ligand interactions analysis.