Utilizing Machine Learning for Efficient Parameterization of Coarse Grained Molecular Force Fields

Data-driven parametrization of molecular mechanics force fields for expansive chemical space coverage

A force field is a critical component in molecular dynamics simulations for computational drug discovery. It must achieve high accuracy within the constraints of molecular mechanics' (MM) limited functional forms, which offers high computational efficiency. With the rapid expansion of synthetically accessible chemical space, traditional look-up table approaches face significant challenges. In this study, we address this issue using a modern data-driven approach, developing ByteFF, an Amber-compatible force field for drug-like molecules. To create ByteFF, we generated an expansive and highly diverse molecular dataset at the B3LYP-D3(BJ)/DZVP level of theory. This dataset includes 2.4 million optimized molecular fragment geometries with analytical Hessian matrices, along with 3.2 million torsion profiles. We then trained an edge-augmented, symmetry-preserving molecular graph neural network (GNN) on this dataset, employing a carefully optimized training strategy. Our model predicts all bonded and non-bonded MM force field parameters for drug-like molecules simultaneously across a broad chemical space. ByteFF demonstrates state-of-the-art performance on various benchmark datasets, excelling in predicting relaxed geometries, torsional energy profiles, and conformational energies and forces. Its exceptional accuracy and expansive chemical space coverage make ByteFF a valuable tool for multiple stages of computational drug discovery.

A Neural-Network-Based Mapping and Optimization Framework for High-Precision Coarse-Grained Simulation

The accuracy and efficiency of a coarse-grained (CG) force field are pivotal for high-precision molecular simulations of large systems with complex molecules. We present an automated mapping and optimization framework for molecular simulation (AMOFMS), which is designed to streamline and improve the force field optimization process. It features a neural-network-based mapping function, DSGPM-TP (deep supervised graph partitioning model with type prediction). This model can accurately and efficiently convert atomistic structures to CG mappings, reducing the need for manual intervention. By integrating bottom-up and top-down methodologies, AMOFMS allows users to freely combine these approaches or use them independently as optimization targets. Moreover, users can select and combine different optimizers to meet their specific mission. With its parallel optimizer, AMOFMS significantly accelerates the optimization process, reducing the time required to achieve optimal results. Successful applications of AMOFMS include parameter optimizations for systems such as POPC and PEO, demonstrating its robustness and effectiveness. Overall, AMOFMS provides a general and flexible framework for the automated development of high-precision CG force fields.

NICE-FF: A non-empirical, intermolecular, consistent, and extensible force field for nucleic acids and beyond

A new non-empirical ab initio intermolecular force field (NICE-FF in buffered 14-7 potential form) has been developed for nucleic acids and beyond based on the dimer interaction energies (IEs) calculated at the spin component scaled-MI-second order Møller-Plesset perturbation theory. A fully automatic framework has been implemented for this purpose, capable of generating well-polished computational grids, performing the necessary ab initio calculations, conducting machine learning (ML) assisted force field (FF) parametrization, and extending existing FF parameters by incorporating new atom types. For the ML-assisted parametrization of NICE-FF, interaction energies of ∼18 000 dimer geometries (with IE < 0) were used, and the best fit gave a mean square deviation of about 0.46 kcal/mol. During this parametrization, atom types apparent in four deoxyribonucleic acid (DNA) bases have been first trained using the generated DNA base datasets. Both uracil and hypoxanthine, which contain the same atom types found in DNA bases, have been considered as test molecules. Three new atom types have been added to the DNA atom types by using IE datasets of both pyrazinamide and 9-methylhypoxanthine. Finally, the last test molecule, theophylline, has been selected, which contains already-fitted atom-type parameters. The performance of NICE-FF has been investigated on the S22 dataset, and it has been found that NICE-FF outperforms the well-known FFs by generating the most consistent IEs with the high-level ab initio ones. Moreover, NICE-FF has been integrated into our in-house developed crystal structure prediction (CSP) tool [called FFCASP (Fast and Flexible CrystAl Structure Predictor)], aiming to find the experimental crystal structures of all considered molecules. CSPs, which were performed up to 4 formula units (Z), resulted in NICE-FF being able to locate almost all the known experimental crystal structures with sufficiently low RMSD20 values to provide good starting points for density functional theory optimizations.

Exploring the role of microbial proteins in controlling environmental pollutants based on molecular simulation

Molecular simulation has been widely used to study microbial proteins' structural composition and dynamic properties, such as volatility, flexibility, and stability at the microscopic scale. Herein, this review describes the key elements of molecular docking and molecular dynamics (MD) simulations in molecular simulation; reviews the techniques combined with molecular simulation, such as crystallography, spectroscopy, molecular biology, and machine learning, to validate simulation results and bridge information gaps in the structure, microenvironmental changes, expression mechanisms, and intensity quantification; illustrates the application of molecular simulation, in characterizing the molecular mechanisms of interaction of microbial proteins with four different types of contaminants, namely heavy metals (HMs), pesticides, dyes and emerging contaminants (ECs). Finally, the review outlines the important role of molecular simulations in the study of microbial proteins for controlling environmental contamination and provides ideas for the application of molecular simulation in screening microbial proteins and incorporating targeted mutagenesis to obtain more effective contaminant control proteins.

Advances in Computational Approaches for Estimating Passive Permeability in Drug Discovery

Passive permeation of cellular membranes is a key feature of many therapeutics. The relevance of passive permeability spans all biological systems as they all employ biomembranes for compartmentalization. A variety of computational techniques are currently utilized and under active development to facilitate the characterization of passive permeability. These methods include lipophilicity relations, molecular dynamics simulations, and machine learning, which vary in accuracy, complexity, and computational cost. This review briefly introduces the underlying theories, such as the prominent inhomogeneous solubility diffusion model, and covers a number of recent applications. Various machine-learning applications, which have demonstrated good potential for high-volume, data-driven permeability predictions, are also discussed. Due to the confluence of novel computational methods and next-generation exascale computers, we anticipate an exciting future for computationally driven permeability predictions.

Quantitative investigation of surfactant monolayer bending tendency at an oil-polar solvent interface using DPD modeling and artificial neural networks

The bending tendency of a surfactant monolayer at an interface is critical in determining the type of emulsion formed and the proximity of the emulsion system to its equilibrium state. Despite its importance, the influence of interaction and surfactant structure on the bending tendency has not been quantitatively investigated. In this study, we develop and validate an artificial neural network (ANN) model based on the torque densities from dissipative particle dynamics (DPD) simulations to address this gap. With the validated ANN model, the relationship between surfactant monolayer bending tendency and all the interaction parameters, oil size, and surfactant structure (size and tail branching) was derived, from which the significance of each factor was ranked. With this ANN model, both the relationship and factor analysis can be instantly investigated without further DPD modeling. Furthermore, we expand the study to surfactant-oil-polar solvent (SOP) systems by varying the interaction parameters between polar solvents (PP). Our finding indicates that the interaction between polar solvents plays an important role in determining the bending tendency of surfactant monolayers; weaker intermolecular attraction between polar solvents makes surfactants tend to bend toward the oil phase (tend to form oil in polar solvent emulsion). Factor analysis reveals that increasing the repulsion between head-head (HH) or head-oil (HO) makes the model surfactants more polar-solvophilic, while increasing the repulsion between polar solvent-head (PH), tail-tail (TT) or oil-oil (OO) makes the model surfactants more lipophilic. The ANN model effectively reproduces the dependence of surfactant monolayer bending tendency on oil size, consistent with experimental observations, the larger the oil size, the higher the bending tendency toward the oil phase. The most intriguing insight derived from the ANN model here is that the effect of branching in the lipophilic tail will be enhanced by factors that make surfactants behave more lipophilic in a surfactant-oil-polar solvent (SOP) system, for rather polar-solvophilic surfactants, the effect of tail branching is negligible.

Machine Learning-Guided Adaptive Parametrization for Coupling Terms in a Mixed United-Atom/Coarse-Grained Model for Diphenylalanine Self-Assembly in Aqueous Ionic Liquids

Precise regulation of the peptide self-assembly into ordered nanostructures with intriguing properties has attracted intense attention. However, predicting peptide assembly at atomic resolution is a challenge due to both the structural flexibility of peptides and the associated huge computational costs. A machine learning-guided adaptive parametrization method was proposed for developing a mixed atomic and coarse-grained (CG) model through a multiobjective optimization strategy. Our model incorporates the united-atom (UA) model for diphenylalanine (P) and the polarizable electrostatic-variable coarse-grained (VaCG) model for aqueous ionic liquid [BMIM]+[BF4]- solution. In this mixed model, the coupling van der Waals (vdW) interaction is addressed by introducing virtual sites (VS) in the UA model to interact with solvent CG beads. The coupling parameters, including the electrostatic parameter and vdW parameters, are automatically optimized through ML-guided adaptive parametrization. The performance of this model was tested by some microstructural properties, e.g., the average number of P-P intermolecular hydrogen bonds (HBs) and radius distribution functions (RDFs) between P and different fragments of IL, in comparison with all-atom (AA) simulations. The computational cost is significantly reduced using such a parametrization scheme, which could search tens of thousands of force-field parameter sets, while needing only a small fraction of them to be assessed with molecular dynamics (MD) simulations. We used such a mixed resolution model to investigate the self-assembly in IL-water mixtures with variants of IL concentration (X). The long-range-ordered fibril structure is formed in a pure water system (X = 0). With an increase of IL concentrations, the formation of an ordered self-assembly nanostructure is prohibited, instead forming branched fibril at X = 2 mol % or amorphous aggregates when X > 10 mol %, resulting from the interplay between π-stacking and HB interactions between P and IL. The qualitative agreement between the simulated structures and the observed morphologies in experiments indicates the applicability of ML-guided parametrization strategy in the study of complex systems, such as polymers, lipid bilayers, and polysaccharides.

Performance efficient macromolecular mechanics via sub-nanometer shape based coarse graining

Dimensionality reduction via coarse grain modeling is a valuable tool in biomolecular research. For large assemblies, ultra coarse models are often knowledge-based, relying on a priori information to parameterize models thus hindering general predictive capability. Here, we present substantial advances to the shape based coarse graining (SBCG) method, which we refer to as SBCG2. SBCG2 utilizes a revitalized formulation of the topology representing network which makes high-granularity modeling possible, preserving atomistic details that maintain assembly characteristics. Further, we present a method of granularity selection based on charge density Fourier Shell Correlation and have additionally developed a refinement method to optimize, adjust and validate high-granularity models. We demonstrate our approach with the conical HIV-1 capsid and heteromultimeric cofilin-2 bound actin filaments. Our approach is available in the Visual Molecular Dynamics (VMD) software suite, and employs a CHARMM-compatible Hamiltonian that enables high-performance simulation in the GPU-resident NAMD3 molecular dynamics engine.

Phase Behavior of Alkyl Ethoxylate Surfactants in a Dissipative Particle Dynamics Model

We present a dissipative particle dynamics (DPD) model capable of capturing the liquid state phase behavior of nonionic surfactants from the alkyl ethoxylate (C_nE_m) family. The model is based upon our recent work [Anderson et al. J. Chem. Phys. 2017, 147, 094503] but adopts tighter control of the molecular structure by setting the bond angles with guidance from molecular dynamics simulations. Changes to the geometry of the surfactants were shown to have little effect on the predicted micelle properties of sampled surfactants, or the water-octanol partition coefficients of small molecules, when compared to the original work. With these modifications the model is capable of reproducing the binary water-surfactant phase behavior of nine surfactants (C₈E₄, C₈E₅, C₈E₆, C₁₀E₄, C₁₀E₆, C₁₀E₈, C₁₂E₆, C₁₂E₈, and C₁₂E₁₂) with a good degree of accuracy.

COGITO: A Coarse-Grained Force Field for the Simulation of Macroscopic Properties of Triacylglycerides

The use of molecular dynamics simulations is becoming ever more widespread; however, the application of this to pure triacylglyceride (TAG) systems is not. In this study, we are presenting the development, and validation, of a new force field (FF), which we have called the COarse-Grained Interchangeable Triacylglyceride-Optimized FF. The FF has been developed using both a bottom-up and top-down approach for different parameters, with the non-bonded parameters being optimized using a Bayesian optimization method. While the FF was developed using monounsaturated TAGs, results show that it is also suitable for fully saturated TAGs. Description of molecules which were not used during the development of the FF is carried out simply by interchanging the bead in the molecule topologies. Results show that the FF can reproduce the macroscopic properties (density and lattice parameters) of pure TAGs as both crystals and melt with high accuracy, as well as reproduce the differences in enthalpies.

Approaches and Perspective of Coarse-Grained Modeling and Simulation for Polymer-Nanoparticle Hybrid Systems

Molecular modeling and simulations have emerged as effective and indispensable tools to characterize polymeric systems. They provide fundamental and essential insights to design a product of the required properties and to improve the understanding of a phenomenon at the molecular level for a particular system. The polymer-nanoparticle hybrids are materials with outstanding properties and correspondingly large applications whose study has benefited from this new paradigm. However, despite the significant expansion of modern day computational powers, investigation of the long time and large length scale phenomenon in polymeric and polymer-nanoparticle systems is still a challenging task to complete through all-atom molecular dynamics (AA-MD) simulations. To circumvent this problem, a variety of coarse-grained (CG) models have been proposed, ranging from the generic CG models for qualitative properties predictions to more realistic chemically specific CG models for quantitative properties predictions. These CG models have already delivered some success stories in the study of several spatial and temporal evolutions of many processes. Some of these studies were beyond the feasibility of traditional atomistic resolution models due to either the size or the time constraints. This review captures the different types of popular CG approaches that are utilized in the investigation of the microscopic behavior of polymer-nanoparticle hybrid systems. The rationale of this article is to furnish an overview of the popular CG approaches and their applications, to review several important and most recent developments, and to delineate the perspectives on future directions in the field.

Molecular Simulation Approaches to the Study of Thermotropic and Lyotropic Liquid Crystals

Over the last decade, the availability of computer time, together with new algorithms capable of exploiting parallel computer architectures, has opened up many possibilities in molecularly modelling liquid crystalline systems. This perspective article points to recent progress in modelling both thermotropic and lyotropic systems. For thermotropic nematics, the advent of improved molecular force fields can provide predictions for nematic clearing temperatures within a 10 K range. Such studies also provide valuable insights into the structure of more complex phases, where molecular organisation may be challenging to probe experimentally. Developments in coarse-grained models for thermotropics are discussed in the context of understanding the complex interplay of molecular packing, microphase separation and local interactions, and in developing methods for the calculation of material properties for thermotropics. We discuss progress towards the calculation of elastic constants, rotational viscosity coefficients, flexoelectric coefficients and helical twisting powers. The article also covers developments in modelling micelles, conventional lyotropic phases, lyotropic phase diagrams, and chromonic liquid crystals. For the latter, atomistic simulations have been particularly productive in clarifying the nature of the self-assembled aggregates in dilute solution. The development of effective coarse-grained models for chromonics is discussed in detail, including models that have demonstrated the formation of the chromonic N and M phases.

Estimating Preferred Alkane Carbon Numbers of Nonionic Surfactants in Normalized Hydrophilic-Lipophilic Deviation Theory from Dissipative Particle Dynamics Modeling

The preferred alkane carbon number (PACN) in the normalized hydrophilic-lipophilic deviation (HLD_N) theory is a numerical parameter and a transferable scale to characterize the amphiphilicity of surfactants, which is usually measured experimentally using the fish diagram or phase inversion temperature (PIT) methods, and the experimental measurement can only be applied to existing surfactants. Here, for the first time, we propose a procedure to estimate the PACN of C_iE_j nonionic surfactants directly from dissipative particle dynamics (DPD) simulation. The procedure leverages the method of moment concept to quantitatively evaluate the bending tendency of nonionic surfactant monolayers by calculating the torque density. Seven nonionic surfactants, C_iE_j (C₆E₂, C₆E₃, C₈E₃, C₈E₄, C₁₀E₄, C₁₂E₄, and C₁₂E₅), with known PACNs are modeled. Two surfactants, C₁₀E₄ and C₆E₂, were first selected to train and test the interaction parameters, and the relationship between interaction parameters and torque density was mapped for the C₁₀E₄-octane-water system using the artificial neural network (ANN) fitting approach to derive the interaction parameters giving zero torque density, then the interaction parameters were tested in the C₆E₂-dodecane-water system to get the final tuned interaction parameters for PACN estimation. With this procedure, we reproduce the PACN values and their trend of seven nonionic surfactants with reasonable accuracy, which opens the door for quantitative comparison of surfactant amphiphilicity and surfactant classification in silico using the PACN as a transferrable scale.

Phase classification using neural networks: application to supercooled, polymorphic core-softened mixtures

Characterization of phases of soft matter systems is a challenge faced in many physical chemical problems. For polymorphic fluids it is an even greater challenge. Specifically, glass forming fluids, as water, can have, besides solid polymorphism, more than one liquid and glassy phases, and even a liquid-liquid critical point. In this sense, we apply a neural network algorithm to analyze the phase behavior of a mixture of core-softened fluids that interact through the continuous-shouldered well (CSW) potential, which have liquid polymorphism and liquid-liquid critical points, similar to water. We also apply the neural network to mixtures of CSW fluids and core-softened alcohols models. We combine and expand methods based on bond-orientational order parameters to study mixtures, applied to mixtures of hardcore fluids and to supercooled water, to include longer range coordination shells. With this, the trained neural network was able to properly predict the crystalline solid phases, the fluid phases and the amorphous phase for the pure CSW and CSW-alcohols mixtures with high efficiency. More than this, information about the phase populations, obtained from the network approach, can help verify if the phase transition is continuous or discontinuous, and also to interpret how the metastable amorphous region spreads along the stable high density fluid phase. These findings help to understand the behavior of supercooled polymorphic fluids and extend the comprehension of how amphiphilic solutes affect the phases behavior.

Machine Learning Directed Optimization of Classical Molecular Modeling Force Fields

Accurate force fields are necessary for predictive molecular simulations. However, developing force fields that accurately reproduce experimental properties is challenging. Here, we present a machine learning directed, multiobjective optimization workflow for force field parametrization that evaluates millions of prospective force field parameter sets while requiring only a small fraction of them to be tested with molecular simulations. We demonstrate the generality of the approach and identify multiple low-error parameter sets for two distinct test cases: simulations of hydrofluorocarbon (HFC) vapor-liquid equilibrium (VLE) and an ammonium perchlorate (AP) crystal phase. We discuss the challenges and implications of our force field optimization workflow.

Gaussian Process Regression for Materials and Molecules

We provide an introduction to Gaussian process regression (GPR) machine-learning methods in computational materials science and chemistry. The focus of the present review is on the regression of atomistic properties: in particular, on the construction of interatomic potentials, or force fields, in the Gaussian Approximation Potential (GAP) framework; beyond this, we also discuss the fitting of arbitrary scalar, vectorial, and tensorial quantities. Methodological aspects of reference data generation, representation, and regression, as well as the question of how a data-driven model may be validated, are reviewed and critically discussed. A survey of applications to a variety of research questions in chemistry and materials science illustrates the rapid growth in the field. A vision is outlined for the development of the methodology in the years to come.

Accelerating computational discovery of porous solids through improved navigation of energy-structure-function maps

While energy-structure-function (ESF) maps are a powerful new tool for in silico materials design, the cost of acquiring an ESF map for many properties is too high for routine integration into high-throughput virtual screening workflows. Here, we propose the next evolution of the ESF map. This uses parallel Bayesian optimization to selectively acquire energy and property data, generating the same levels of insight at a fraction of the computational cost. We use this approach to obtain a two orders of magnitude speedup on an ESF study that focused on the discovery of molecular crystals for methane capture, saving more than 500,000 central processing unit hours from the original protocol. By accelerating the acquisition of insight from ESF maps, we pave the way for the use of these maps in automated ultrahigh-throughput screening pipelines by greatly reducing the opportunity risk associated with the choice of system to calculate.

Critical Micelle Concentrations in Surfactant Mixtures and Blends by Simulation

We explore the use of coarse-grained dissipative particle dynamics simulations to predict critical micelle concentrations (CMCs) in polydisperse surfactant mixtures and blends. By fitting pseudo-phase separation models (PSMs) to aqueous solutions of binary surfactant mixtures at selected compositions above the CMC, we avoid the need for expensive simulations of more complex multicomponent mixtures performed as a function of dilution. The approach is demonstrated for sodium laureth sulfate (SLES) surfactants with polydispersity in the ethoxylate spacer. For this system, we find a modest degree of cooperativity in micelle formation, which we attribute to the reduced repulsion between charged headgroups for surfactants with dissimilar ethoxylate spacer lengths. However, this is insufficient to explain the lowered CMC often observed in commercial SLES samples, which we attribute to the presence of small amounts of unsulfated alkyl ethoxylates and/or traces of salt.

Recent Open Issues in Coarse Grained Force Fields

This viewpoint intends to show recent open issues of using coarse grained models in molecular dynamics simulation. It reviews the current knowledge of the comparison between experimental and simulation data of structural and physical chemical properties that depend on the hydrophilic and hydrophobic behavior of the molecule.

Wax Formation in Linear and Branched Alkanes with Dissipative Particle Dynamics

We present a dissipative particle dynamics (DPD) model for wax formation (i.e., the freezing transition) in linear and branched alkanes at room temperature (298 K) and atmospheric pressure. We parametrize the model using pure liquid phase densities and the onset of wax formation as a function of alkyl chain length. Significant emphasis is placed on building an accurate representation of the underlying molecular architecture by careful consideration of bond lengths and angles, aided by distributions obtained from molecular dynamics simulation. Using the derived model, we observe wax formation in n-alkanes when the alkyl chain length is greater than 18 (n-octadecane), in excellent agreement with experimental observations. Further, we reproduce the behavior of branched alkanes and mixtures including solubilities of heavy alkanes in light alkane solvents.

Model for the Simulation of the CnEm Nonionic Surfactant Family Derived from Recent Experimental Results

Using a comprehensive set of recently published experimental results for training and validation, we have developed computational models appropriate for simulations of aqueous solutions of poly(ethylene oxide) alkyl ethers, an important class of micelle-forming nonionic surfactants, usually denoted C_nE_m. These models are suitable for use in simulations that employ a moderate amount of coarse graining and especially for dissipative particle dynamics (DPD), which we adopt in this work. The experimental data used for training and validation were reported earlier and produced in our laboratory using dynamic light scattering (DLS) measurements performed on 12 members of the C_nE_m compound family yielding micelle size distribution functions and mass-weighted mean aggregation numbers at each of several surfactant concentrations. The range of compounds and quality of the experimental results were designed to support the development of computational models. An essential feature of this work is that all simulation results were analyzed in a way that is consistent with the experimental data. Proper account is taken of the fact that a broad distribution of micelle sizes exists, so mass-weighted averages (rather than number-weighted averages) over this distribution are required for the proper comparison of simulation and experimental results. The resulting DPD force field reproduces several important trends seen in the experimental critical micelle concentrations and mass-averaged mean aggregation numbers with respect to surfactant characteristics and concentration. We feel it can be used to investigate a number of open questions regarding micelle sizes and shapes and their dependence on surfactant concentration for this important class of nonionic surfactants.