A Reliable Docking/Scoring Scheme Based on the Semiempirical Quantum Mechanical PM6-DH2 Method Accurately Covering Dispersion and H-Bonding: HIV-1 Protease with 22 Ligands

Abstract

In this study, we introduce a fast and reliable rescoring scheme for docked complexes based on a semiempirical quantum mechanical PM6-DH2 method. The method utilizes a PM6-based Hamiltonian with corrections for dispersion energy and hydrogen bonds. The total score is constructed as the sum of the PM6-DH2 interaction enthalpy, the empirical force field (AMBER) interaction entropy, and the sum of the deformation (PM6-DH2, SMD) and the desolvation (SMD) energies of the ligand. The main advantage of the procedure is the fact that we do not add any empirical parameter for either an individual component of the total score or an individual protein-ligand complex. This rescoring method is applied to a very challenging system, namely, the HIV-1 protease with a set of ligands. As opposed to the conventional DOCK procedure, the PM6-DH2 rescoring based on all of the terms distinguishes between binders and nonbinders and provides a reliable correlation of the theoretical and experimental binding free energies. Such a dramatic improvement, resulting from the PM6-DH2 rescoring of all the complexes, provides a valuable yet inexpensive tool for rational drug discovery and de novo ligand design.

The Semiempirical Quantum Mechanical Scoring Function for In Silico Drug Design

“…[3] These quantities are not only difficult to calculate but also at least one order of magnitude larger than the experimental absolute binding free energy. [69] Even worse, the difference between a tight-binding inhibitor (low nanomolar activity) and the weakest experimentally detectable ligand (an activity of about 100 mm), that is, the so-called "window of activity" is only about 5 kcal mol À1 . [70] Bearing in mind that some of the terms listed above harbour an uncertainty of AE 1-2 kcal mol À1 , the error propagation may lead to a total uncertainty of approximately 5 kcal mol À1 .…”

“…[78] The SQM scoring was developed further in our group by combining the advanced empirical corrections to the SQM methods (see above) with a refined QM-based scoring function. [69] Like the previous approaches, [79,80] our QM scoring function is designed as a sum of the gas-phase interaction energy, the solvation/desolvation free energy, the change of the conformational "free" energies of the protein and ligand and the entropy change upon binding. Such a decomposition of the binding free energy (approximated by the score) reflects the phenomenological decomposition of an idealised binding process (Figure 3).…”

A semiempirical approach to ligand‐binding affinities: Dependence on the Hamiltonian and corrections

“…17,18,19 They also employ the more recent PM6 20 method with corrections for dispersion and hydrogen-bond interactions (DH2). 21,22 They use a single minimised structure of the complex, 3 although they have shown that energies calculated from a minimised structure of the ligand deviates by 13 kJ/mol on average from those calculated on an ensemble of structures from a MD simulation.…”

Empirical Hydrogen‐Bond Potential Functions—An Old Hat Reconditioned

“…[72,73] Most interesting are probably the attempts by Hobza and co-workers to develop a reliable docking and scoring scheme based on PM6-DH2, [74,75] and their straight-forward extension of the DH2 model to halogen bonds. [76] One important problem specific to the DH2 model (and other models utilizing charges, such as DH1 and FS1) is the need for charge derivatives to calculate analytical gradients, which is ignored in all current DH2 implementations, but can be quite substantial in the case of strong H bonds.…”