Generation of Quantum Configurational Ensembles Using Approximate Potentials

Abstract

Conformational analysis is of paramount importance in drug design: it is crucial to determine pharmacological properties, understand molecular recognition processes, and characterize the conformations of ligands when unbound. Molecular Mechanics (MM) simulation methods, such as Monte Carlo (MC) and molecular dynamics (MD), are usually employed to generate ensembles of structures due to their ability to extensively sample the conformational space of molecules. The accuracy of these MM-based schemes strongly depends on the functional form of the force field (FF) and its parametrization, components that often hinder their performance. High-level methods, such as ab initio MD, provide reliable structural information but are still too computationally expensive to allow for extensive sampling. Therefore, to overcome these limitations, we present a multilevel MC method that is capable of generating quantum configurational ensembles while keeping the computational cost at a minimum. We show that FF reparametrization is an efficient route to generate FFs that reproduce QM results more closely, which, in turn, can be used as low-cost models to achieve the gold standard QM accuracy. We demonstrate that the MC acceptance rate is strongly correlated with various phase space overlap measurements and that it constitutes a robust metric to evaluate the similarity between the MM and QM levels of theory. As a more advanced application, we present a self-parametrizing version of the algorithm, which combines sampling and FF parametrization in one scheme, and apply the methodology to generate the QM/MM distribution of a ligand in aqueous solution.

Does a Machine-Learned Potential Perform Better Than an Optimally Tuned Traditional Force Field? A Case Study on Fluorohydrins

“…Since we obtained hMC acceptance rates of ≥90% for all molecules in the test set (Figure ), we conclude that ANI-2x is a viable model to use in MD simulations. These high hMC acceptance rates are comparable in magnitude to those we obtained in a previous study of molecules of similar size modeled using GAFF-like FFs . On the other hand, the ANI-2x to DFT acceptance rate measures the similarity between the ANI-2x potential and the ωB97X/6-31G* level of theory.…”

“…The nMC-MC simulations aimed at generating ensembles of configurations representative of the ωB97X/6-31G* ensemble by using ANI-2x as the approximate potential. The acceptance rates that the nMC-MC simulations produce give important information about the ANI-2x performance in the gas phase . On the one hand, the hMC acceptance rate measures the stability of the short MD runs performed in the nMC-MC algorithm, and it is positively correlated with energy conservation during the MD run.…”

“…The acceptance rates that the nMC-MC simulations produce give important information about the ANI-2x performance in the gas phase. 21 On the one hand, the hMC acceptance rate measures the stability of the short MD runs performed in the nMC-MC algorithm, and it is positively correlated with energy conservation during the MD run. Since we obtained hMC acceptance rates of ≥90% for all molecules in the test set ( Figure 3 ), we conclude that ANI-2x is a viable model to use in MD simulations.…”

“…This training data set was composed of structures sampled from the DFT ensemble using the nested Markov chain Monte Carlo (nMC-MC) algorithm , implemented in ParaMol, interfaced with Psi4 for the QM calculations. nMC-MC combines sampling at an approximate potential with periodic switching attempts to the QM level, enabling recovery of the exact quantum DFT ensemble . Because of this feature, this method was used to generate high-quality structures representative of our reference level of theory.…”

“…nMC-MC combines sampling at an approximate potential with periodic switching attempts to the QM level, enabling recovery of the exact quantum DFT ensemble. 21 Because of this feature, this method was used to generate high-quality structures representative of our reference level of theory. For each γ-fluorohydrin, a variable number of nMC-MC samplers were spawned, each starting from the major conformers reported in ref ( 38 ) (the same conformers used to calculate the ESPs for the RESP procedure).…”