An attempt to incorporate effect of direct interaction between a ligand and explicit water molecules into MM/3D‐RISM

Abstract

Endpoint methods using continuum-solvent models are widely used to estimate protein-ligand affinity. A recently developed method, MM/3D-RISM, estimates the solvation energy using statistical mechanics by 3D-RISM. This method is theoretically expected to accurately describe solvation effects and to also be less dependent on protein-ligand systems. In this study, we examined the performance of MM/3D-RISM for a set of α-thrombin inhibitors with a non-congeneric series of ligands, containing three diverse chemical scaffolds. The standard MM/3D-RISM showed a weak correlation (R = 0.191) but correctly estimated affinity for two of the three scaffolds. However, the simplest inhibitor, benzamidine, was not ranked appropriately. From visual inspection of inhibitor-binding modes, an attempt was made to incorporate the direct interaction between a ligand and water molecules into MM/3D-RISM. A model (Model-1) dealing with directly interacting water molecules (Wat) as an independent component of a protein (R)-ligand (L) complex-formation, that is, R + L + Wat → R-L-Wat, showed a better linearity (R = 0.422) than that of the standard MM/3D-RISM model and achieved a good ranking of all three scaffolds of α-thrombin inhibitors. Additionally, an attempt was made to model avidin-biotin system with a congeneric series of inhibitors, and results showed that both the standard MM/3D-RISM model (R = 0.839) and Model-1 (R = 0.695) satisfactorily estimated the affinity.

3D-RISM-AI: A Machine Learning Approach to Predict Protein–Ligand Binding Affinity Using 3D-RISM

“…Recently, instead of continuum models, a solution statistical-mechanics theory-based method, called the three-dimensional reference interaction model (3D-RISM), was combined with MM. 11 − 16 The 3D-RISM method can evaluate HFEs, preserving the molecular features. 34 − 37 The 3D-RISM method provides a thermally averaged distribution of solvent molecules around solute molecules, and HFE can be calculated through an integral equation using the distribution function.…”

“… 40 The 3D-RISM method, combined with MM or extended protocols described above, improves the prediction of HFEs. 11 − 16 , 38 − 40 …”

“…Various computational methods for BFE calculation based on physicochemical approaches ,,− using the thermodynamic cycle (Figure ) or data-driven approaches using machine learning − have been proposed to date. The required computational burdens of physicochemical approaches differ depending on the incorporation methods of the hydration effects.…”

“…In molecular mechanics (MM) and Poisson–Boltzmann surface area (PB/SA) or MM and generalized Born surface area (GB/SA) methods, the interaction energy between a protein and a ligand and the configurational entropies are calculated using MM, where the hydration free energies (HFEs) of the protein, ligand, and complex are approximately calculated using a continuous dielectric model (i.e., implicit solvent model). , In continuous dielectric models, the detailed molecular features of water molecules such as hydrogen bonding, the hydrophobic effect, and the rearrangement of the water molecules upon solute insertion are missing. Recently, instead of continuum models, a solution statistical-mechanics theory-based method, called the three-dimensional reference interaction model (3D-RISM), was combined with MM. − The 3D-RISM method can evaluate HFEs, preserving the molecular features. − The 3D-RISM method provides a thermally averaged distribution of solvent molecules around solute molecules, and HFE can be calculated through an integral equation using the distribution function. The MM/3D-RISM method can successfully predict the BFE of a group of ligands with different activities for the same protein.…”

“…The 3D-RISM method has also been applied to the statistical analysis of hydration water based on a large number of 3D-RISM calculations, development of an efficient method to calculate HFEs, or binding-site searches by extending the theory to the distribution of atoms constituting the ligand . The 3D-RISM method, combined with MM or extended protocols described above, improves the prediction of HFEs. − ,− …”