Molecular Modelling of Ionic Liquids: Force-Field Validation and Thermodynamic Perspective from Large-Scale Fast-Growth Solvation Free Energy Calculations

Molecular simulations are becoming a common tool for the investigation of dynamic and thermodynamic properties of novel solvents such as ionic liquids and the more recent deep eutectic solvents. As the electrostatics derived from ab initio calculations often fail to reproduce the experimental behaviors of these functionalized solvents, a common treatment is scaling the atomic charges to improve the accord between experimental and computational results for some selected properties, e.g., the density of the liquids. Although there are many computational benchmarks on structural properties of bulk ionic liquids, the choice of the best scaling parameter remains an open question. As these liquids are designed to solvate solutes, whether the solvation thermodynamics could be correctly described is of utmost importance in practical situations. Therefore, in the current work, we provide a thermodynamic perspective of this charge scaling issue directly from solute-solvent interactions. We present a comprehensive large-scale calculation of solvation free energies via nonequilibrium fast-switching simulations for a spectrum of molecules in ionic liquids, the atomic charges of which derived from ab initio calculations are scaled to find the best scaling factor that maximizes the prediction-experiment correlation. The density-derived choice of the scaling parameter as the estimate from bulk properties is compared with the solvation-free-energy-derived one. We observed that when the scaling factor is decreased from 1.0 to 0.5, the mass density exhibits a monotonically decreasing behavior, which is caused by weaker inter-molecular interactions produced by the scaled atomic charges. However, the solvation free energies of external agents do not show consistent monotonic behaviors like the bulk property, the underlying physics of which are elucidated to be the competing electrostatic and vdW responses to the scaling-parameter variation. More intriguingly, although the recommended value for charge scaling from bulk properties falls in the neighborhood of 0.6~0.7, solvation free energies calculated at this value are not in good agreement with the experimental reference. By modestly increasing the scaling parameter (e.g., by 0.1) to avoid over-scaling of atomic charges, the solute-solvent interaction free energy approaches the reference value and the quality of calculated solvation thermodynamics approaches the hydration case. According to this phenomenon, we propose a feasible way to obtain the best scaling parameter that produces balanced solute-solvent and solvent-solvent interactions, i.e., first scanning the density-scaling-factor profile and then adding ~0.1 to that solution. We further calculate the partition coefficient or transfer free energy of solutes from water to ionic liquids to provide another thermodynamic perspective of the charge scaling benchmark. Another central result of the current work is about the widely used force fields to describe bonded and vdW terms for ionic liquids derivatives. These pre-fitted transferable parameters are evaluated and refitted in a system-specific manner to provide a detailed assessment of the reliability and accuracy of these commonly used parameters. Component-specific refitting procedures unveil that the bond-stretching term is the most problematic part of the GAFF derivatives and the angle-bending term in some cases is also not accurate enough. Astonishingly, the torsional potential defined in these pre-fitted force fields performs extremely well.

Implicitly perturbed Hamiltonian: a class of versatile and general-purpose molecular representations for machine learning

Unraveling challenging problems by machine learning has recently become a hot topic in many scientific disciplines. For developing rigorous machine-learning models to study problems of interest in molecular sciences, translating molecular structures to quantitative representations as suitable machine-learning inputs play a central role. Many different molecular representations and the state-of-the-art ones, although efficient in studying numerous molecular features, still are suboptimal in many challenging cases, as discussed in the context of the present research. The main aim of the present study is to introduce the Implicitly Perturbed Hamiltonian (ImPerHam) as a class of versatile representations for more efficient machine learning of challenging problems in molecular sciences. ImPerHam representations are defined as energy attributes of the molecular Hamiltonian, implicitly perturbed by a number of hypothetic or real arbitrary solvents based on continuum solvation models. We demonstrate the outstanding performance of machine-learning models based on ImPerHam representations for three diverse and challenging cases of predicting inhibition of the CYP450 enzyme, high precision, and transferrable evaluation of conformational energy of molecular systems, and accurately reproducing solvation free energies for large benchmark sets.

Implicitly perturbed Hamiltonian: a class of versatile and general-purpose molecular representations for machine leaning

Unraveling challenging problems by machine learning has recently become a hot topic in many scientific disciplines. For developing rigorous machine-learning models to study problems of interest in molecular sciences, translating molecular structures to quantitative representations as suitable machine-learning inputs plays a central role. Many different molecular representations and the state-ofthe- art ones, although efficient in studying numerous molecular features, still are sub-optimal in many challenging cases, as discussed in the context of present research. The main aim of the present study is to introduce the Implicitly Perturbed Hamiltonian (ImPerHam) as a class of versatile representations for more efficient machine learning of challenging problems in molecular sciences. ImPerHam representations are defined as energy attributes of the molecular Hamiltonian, implicitly perturbed by a number of hypothetic or real arbitrary solvents based on continuum solvation models. We demonstrate outstanding performance of machine-learning models based on ImPerHam representations for three diverse and challenging cases of predicting inhibition of the CYP450 enzyme, high precision and transferrable evaluation of conformational energy of molecular systems and accurately reproducing solvation free energies for large benchmark sets.

Improved prediction of solvation free energies by machine-learning polarizable continuum solvation model

Theoretical estimation of solvation free energy by continuum solvation models, as a standard approach in computational chemistry, is extensively applied by a broad range of scientific disciplines. Nevertheless, the current widely accepted solvation models are either inaccurate in reproducing experimentally determined solvation free energies or require a number of macroscopic observables which are not always readily available. In the present study, we develop and introduce the Machine-Learning Polarizable Continuum solvation Model (ML-PCM) for a substantial improvement of the predictability of solvation free energy. The performance and reliability of the developed models are validated through a rigorous and demanding validation procedure. The ML-PCM models developed in the present study improve the accuracy of widely accepted continuum solvation models by almost one order of magnitude with almost no additional computational costs. A freely available software is developed and provided for a straightforward implementation of the new approach.

Optimized Halogen Atomic Radii for PBSA Calculations Using Off-Center Point-Charges

<div>In force field methods, the usage of off-center point-charges, also called extra-points (EPs), is a common strategy to tackle the anisotropy of the electrostatic potential of covalently-bonded halogens (X), thus allowing the description of halogen bonds (XBs) at the molecular mechanics / molecular dynamics (MM/MD) level. Diverse EP implementations exist in the literature differing on the charge sets and/or the X–EP distances. Poisson–Boltzmann and surface area (PBSA) calculations can be used to obtain solvation free energies (∆G solv ) of small molecules, often to compute binding free energies (∆G bind ) at the MM PBSA level. This method depends, among other parameters, on the empirical assignment of atomic radii (PB radii). Given the multiplicity of off-center point-charges models and the lack of specific PB radii for halogens compatible with such implementations, in this work we assessed the performance of PBSA calculations for the estimation of ∆G solv values in water (∆G hyd ), also conducting an optimization of the halogen PB radii (Cl, Br, and I) for each EP model. We not only expand the usage of EP models in the scope of the General AMBER Force Field (GAFF) but also provide the first optimized halogen PB radii in the context of the CHARMM General Force Field (CGenFF), thus contributing to improving the description of halogenated compounds in PBSA calculations.</div>