Using the Semiempirical Quantum Mechanics in Improving the Molecular Docking: A Case Study with CDK2

Abstract

In this study, we use some modified semiempirical quantum mechanics (SQM) methods for improving the molecular docking process. To this end, the three popular SQM Hamiltonians, PM6, PM6‐D3H4X, and PM7 are employed for geometry optimization of some binding modes of ligands docked into the human cyclin‐dependent kinase 2 (CDK2) by two widely used docking tools, AutoDock and AutoDock Vina. The results were analyzed with two different evaluation metrics: the symmetry‐corrected heavy‐atom RMSD and the fraction of recovered ligand‐protein contacts. It is shown that the evaluation of the fraction of recovered contacts is more useful to measure the similarity between two structures when interacting with a protein. It was also found that AutoDock is more successful than AutoDock Vina in producing the correct ligand poses (RMSD≤2.0 Å) and ranking of the poses. It is also demonstrated that the ligand optimization at the SQM level improves the docking results and the SQM structures have a significantly better fit to the observed crystal structures. Finally, the SQM optimizations reduce the number of close contacts in the docking poses and successfully remove most of the clash or bad contacts between ligand and protein.

Use of the Complementarity Principle in Docking Procedures: A New Approach for Evaluating the Correctness of Binding Poses

“…Likewise, basing the ligand–receptor interaction analysis entirely on the docking results can also bring about wrong conclusions. One way to increase the accuracy of the docking results can be to perform geometry optimization of the complex, as was done in the case of CDK2 inhibitors by Bagheri et al 15 This allows a recovery of some ligand–receptor interactions but comes at the expense of computation time. However, it has to be said that poses obtained by docking in this manner are still good starting structures for molecular dynamics (MD) simulations, as the first step of MD simulations is geometry optimization, which “corrects” these initial poses.…”

“…This contribution to the RMSD of substituent groups, which do not significantly interact with receptor atoms, poses a problem in docking validation if docking results are judged solely on the RMSD score. One way to tackle this problem, as it was done by Bagheri et al 15 and Feinstein and Brylinski, 16 is by using the fraction of recovered ligand–receptor contacts, which was defined in a following way: “the contacts were identified for interatomic distances less than 4.5 Å between any pair of heavy atoms, one from the ligand and one from the receptor. The difference of less than 1.0 Å between the predicted contact and the corresponding contact in the experimental structure was considered as a correct recovery of the contact.” Analogously, the similarity of the regression lines obtained by using eq 1 and the AlteQ method for the docked and minimized poses can be regarded as an alternative method of checking the recovered ligand–receptor contacts.…”

Ensemble learning from ensemble docking: revisiting the optimum ensemble size problem

“…A list of these ligands can be found in Table S3 . In a recent study, we compared AutoDock4 9 and AutoDock Vina 10 docking tools regarding their pose prediction accuracy based on available X-ray structures of ligand-CDK2 complexes 64 . It was shown that for the top-ranked predicted pose, i.e., the best scored docking geometry, AutoDock4 reproduced 62% of binding geometries with an RMSD less than 2 Å from the experimental geometry, while this value was 37% for AutoDock Vina.…”

“…Accordingly, AutoDock4 was used in the current study since better pose prediction in the ensemble docking stage can enhance the subsequent ensemble learning results. The docking parameters were chosen to be the same as those used in a previous study 64 . A Lamarckian genetic algorithm with an initial population of 500 was repeated 200 times for each ligand-receptor complex, and the best scored binding mode was selected for subsequent machine learning steps.…”

Identification of Catechins Binding Pockets in Monomeric Aβ42Through Ensemble Docking and MD Simulations

“…Very recently, it has been shown that the correct pose (the pose with the lowest RMSD from the corresponding experimental pose) is usually predicted by Vina but sometimes, does not get the top score in the Vina ranking. [92][93] To avoid the problem and to capture the correct poses, it is recommended that except for the top-ranked pose, some important lower ranked poses for each docking run should be identified and selected for post-docking analysis. For more discussions about ranking, see also Refs.…”