Inertial likelihood maximization for reaction coordinates with high transmission coefficients

Data-Driven Path Collective Variables

Identifying optimal collective variables to model transformations using atomic-scale simulations is a long-standing challenge. We propose a new method for the generation, optimization, and comparison of collective variables that can be thought of as a data-driven generalization of the path collective variable concept. It consists of a kernel ridge regression of the committor probability, which encodes a transformation's progress. The resulting collective variable is one-dimensional, interpretable, and differentiable, making it appropriate for enhanced sampling simulations requiring biasing. We demonstrate the validity of the method on two different applications: a precipitation model and the association of Li+ and F- in water. For the former, we show that global descriptors such as the permutation invariant vector allow reaching an accuracy far from the one achieved via simpler, more intuitive variables. For the latter, we show that information correlated with the transformation mechanism is contained in the first solvation shell only and that inertial effects prevent the derivation of optimal collective variables from the atomic positions only.

An exploration of machine learning models for the determination of reaction coordinates associated with conformational transitions

Determining collective variables (CVs) for conformational transitions is crucial to understanding their dynamics and targeting them in enhanced sampling simulations. Often, CVs are proposed based on intuition or prior knowledge of a system. However, the problem of systematically determining a proper reaction coordinate (RC) for a specific process in terms of a set of putative CVs can be achieved using committor analysis (CA). Identifying essential degrees of freedom that govern such transitions using CA remains elusive because of the high dimensionality of the conformational space. Various schemes exist to leverage the power of machine learning (ML) to extract an RC from CA. Here, we extend these studies and compare the ability of 17 different ML schemes to identify accurate RCs associated with conformational transitions. We tested these methods on an alanine dipeptide in vacuum and on a sarcosine dipeptoid in an implicit solvent. Our comparison revealed that the light gradient boosting machine method outperforms other methods. In order to extract key features from the models, we employed Shapley Additive exPlanations analysis and compared its interpretation with the "feature importance" approach. For the alanine dipeptide, our methodology identifies ϕ and θ dihedrals as essential degrees of freedom in the C7ax to C7eq transition. For the sarcosine dipeptoid system, the dihedrals ψ and ω are the most important for the cisαD to transαD transition. We further argue that analysis of the full dynamical pathway, and not just endpoint states, is essential for identifying key degrees of freedom governing transitions.

ATESA: An Automated Aimless Shooting Workflow

Transition path sampling methods are powerful tools for studying the dynamics of rare events in molecular simulations. However, these methods are generally restricted to experts with the knowledge and resources to properly set up and analyze the often hundreds of thousands of simulations that constitute a complete study. Aimless Transition Ensemble Sampling and Analysis (ATESA) is a new open-source software program written in Python that automates a full transition path sampling workflow based on the aimless shooting algorithm, streamlining the process and reducing the barrier to use for researchers new to this approach. This introduction to ATESA includes a demonstration of a complete transition path sampling process flow for an example reaction, including finding an initial transition state, sampling with aimless shooting, building a reaction coordinate with inertial likelihood maximization, verifying that coordinate with committor analysis, and measuring the reaction energy profile with umbrella sampling. We also describe our implementation of a termination criterion for aimless shooting based on the Godambe information calculated during model building with likelihood maximization as well as a novel approach to constraining simulations to the desired rare event pathway during umbrella sampling.

Nucleation Rate Theory for Coordination Number: Elucidating Water-Mediated Formation of a Zigzag Na2SO4 Morphology

Predicting and controlling nanostructure formation during nucleation can pave the way to synthesizing novel energy materials via crystallization. However, such control over nucleation and crystallization remains challenging due to an inadequate understanding of critical factors that govern evolving atomistic structures and dynamics. Herein, we utilize coordination number as a reaction coordinate and rate theory to investigate how sodium sulfate, commonly known as a phase-change energy material, nucleates in a supersaturated aqueous solution. In conjunction with ab initio and force field-based molecular dynamics simulation, the rate theoretical analysis reveals that sodium sulfate from an initially dissolved metastable state transits to a heterogeneous mixture of prenucleated clusters and finally to a large cylindrical zigzag morphology. Measurements of Raman spectra and their ab initio modeling confirm that this nucleated morphology contains a few waters for every sulfate. Rate processes such as solvent exchange and desolvation exhibit high sensitivity to the evolving prenucleation/nucleation structures, providing a means to distinguish between critical nucleation precursors. Desolvation and forming the first-shell interionic coordination structure via monomer-by-monomer addition around sulfates are found to explain the formation of large nuclei. Thus, a detailed understanding of the step-by-step structure formation across scales has been achieved. This can be leveraged to predict nucleation-related structures and dynamics and potentially control the synthesis of novel phase-change materials for energy applications.

Locating Transition Zone in Phase Space

Understanding the reaction mechanism is required for better control of chemical reactions and is usually achieved by locating transition states (TSs) along a proper one-dimensional coordinate called reaction coordinate (RC). The identification of RC can be very difficult for high-dimensional realistic systems. A number of methods have been proposed to tackle this problem. A machine learning method is developed here to incorporate the influence of velocity on the reaction process. The method is also free of the unbalanced label problem resulting from the rather low fraction of configurations near the TS and can be easily extended to large systems. It locates the transition zone in the phase space and defines the dividing surface with a high transmission coefficient. Moreover, considering that the reaction environment can not only change the reaction path but also activate the reactive mode through energy transfer, we devise two measures to quantify the influence of these two factors on the reaction process and find that solvents can assist the reaction by directly doing work along the reactive mode. Not surprisingly, there is a positive correlation between the efficiency of energy transfer into the reactive mode and the reaction rate.

Explaining reaction coordinates of alanine dipeptide isomerization obtained from deep neural networks using Explainable Artificial Intelligence (XAI)

A method for obtaining appropriate reaction coordinates is required to identify transition states distinguishing product and reactant in complex molecular systems. Recently, abundant research has been devoted to obtaining reaction coordinates using artificial neural networks from deep learning literature, where many collective variables are typically utilized in the input layer. However, it is difficult to explain the details of which collective variables contribute to the predicted reaction coordinates owing to the complexity of the nonlinear functions in deep neural networks. To overcome this limitation, we used Explainable Artificial Intelligence (XAI) methods of the Local Interpretable Model-agnostic Explanation (LIME) and the game theory-based framework known as Shapley Additive exPlanations (SHAP). We demonstrated that XAI enables us to obtain the degree of contribution of each collective variable to reaction coordinates that is determined by nonlinear regressions with deep learning for the committor of the alanine dipeptide isomerization in vacuum. In particular, both LIME and SHAP provide important features to the predicted reaction coordinates, which are characterized by appropriate dihedral angles consistent with those previously reported from the committor test analysis. The present study offers an AI-aided framework to explain the appropriate reaction coordinates, which acquires considerable significance when the number of degrees of freedom increases.

Nonlinear Reaction Coordinate of an Enzyme Catalyzed Proton Transfer Reaction

We present an in-depth study on the theoretical calculation of an optimum reaction coordinate as a linear or nonlinear combination of important collective variables (CVs) sampled from an ensemble of reactive transition paths for an intramolecular proton transfer reaction catalyzed by the enzyme human carbonic anhydrase (HCA) II. The linear models are optimized by likelihood maximization for a given number of CVs. The nonlinear models are based on an artificial neural network with the same number of CVs and optimized by minimizing the root-mean-square error in comparison to a training set of committor estimators generated for the given transition. The nonlinear reaction coordinate thus obtained yields the free energy of activation and rate constant as 9.46 kcal mol^-1 and 1.25 × 10⁶ s^-1, respectively. These estimates are found to be in quantitative agreement with the known experimental results. We have also used an extended autoencoder model to show that a similar analysis can be carried out using a single CV only. The resultant free energies and kinetics of the reaction slightly overestimate the experimental data. The implications of these results are discussed using a detailed microkinetic scheme of the proton transfer reaction catalyzed by HCA II.

A matrix completion algorithm to recover modes orthogonal to the minimum energy path in chemical reactions

Structured statistical methods are promising for recovering or completing information from noisy and incomplete data with high fidelity. In particular, matrix completion exploits underlying structural properties such as rank or sparsity. Our objective is to employ matrix completion to reduce computational effort associated with the calculation of multiple quantum chemical Hessians, which are necessary for identification of temperature-dependent free energy maxima under canonical variational transition state theory (VTST). We demonstrate proof-of-principle of an algebraic variety-based matrix completion method for recovering missing elements in a matrix of transverse Hessian eigenvalues constituting the minimum energy path (MEP) of a reaction. The algorithm, named harmonic variety-based matrix completion (HVMC), utilizes the fact that the points lying on the MEP of a reaction step constitute an algebraic variety in the reaction path Hamiltonian representation. We demonstrate that, with as low as 30% random sampling of matrix elements for the largest system in our test set (46 atoms), the complete matrix of eigenvalues can be recovered. We further establish algorithm performance for VTST rate calculations by quantifying errors in zero-point energies and vibrational free energies. Motivated by this success, we outline next steps toward developing a practical HVMC algorithm, which utilizes a gradient-based sampling protocol for low-cost VTST rate computations.

Hydration structure and water exchange kinetics at xenotime-water interfaces: implications for rare earth minerals separation

Hydration of surface ions gives rise to structural heterogeneity and variable exchange kinetics of water at complex mineral-water interfaces. Here, we employ ab initio molecular dynamics (AIMD) simulations and water adsorption calorimetry to examine the aqueous interfaces of xenotime, a phosphate mineral that contains predominantly Y3+ and heavy rare earth elements. Consistent with natural crystal morphology, xenotime is predicted to have a tetragonal prismatic shape, dominated by the {100} surface. Hydration of this surface induces multilayer interfacial water structures with distinct OH orientations, which agrees with recent crystal truncation rod measurements. The exchange kinetics between two adjacent water layers exhibits a wide range of underlying timescales (5-180 picoseconds), dictated by ion-water electrostatics. Adsorption of a bidentate hydroxamate ligand reveals that {100} xenotime surface can only accommodate monodentate coordination with water exchange kinetics strongly depending on specific ligand orientation, prompting us to reconsider traditional strategies for selective separation of rare-earth minerals.

Structure and dynamics of the molten alkali-chloride salts from an X-ray, simulation, and rate theory perspective

Molten salts are of great interest as alternative solvents, electrolytes, and heat transfer fluids in many emerging technologies.

Orthogonal order parameters to model the reaction coordinate of an enzyme catalyzed reaction

The choice of suitable collective variables in formulating an optimal reaction coordinate is a challenging task for activated transitions between a pair of stable states especially when dealing with biochemical changes such as enzyme catalyzed reactions. A detailed benchmarking study is carried out on the choice of collective variables that can distinguish between the stable states unambiguously. We specifically address the issue if these variables may be directly used to model the optimal reaction coordinate, or if it would be better to use their orthogonalized counterparts. The proposed computational scheme is applied to the rate determining intramolecular proton transfer step in the enzyme human carbonic anhydrase II. The optimum reaction coordinate is determined with and without orthogonalization of the collective variables pertinent to a key conformational fluctuation and the actual proton transfer step at the active site of the enzyme. Suitability of the predicted reaction coordinates in different processes is examined in terms of the free energy profile projected along the reaction coordinate, the rate constant of transition and the underlying molecular mechanism of barrier crossing. Our results indicate that a better agreement with earlier simulation and experimental data is obtained when the orthogonalized collective variables are used to model the reaction coordinate.

Mechanism of oligosaccharide synthesis via a mutant GH29 fucosidase

First unbiased transition path sampling study of a glycosynthase enzyme reveals single-step mechanism with oxocarbenium-like transition state.

Equipartition terms in transition path ensemble: Insights from molecular dynamics simulations of alanine dipeptide

Transition path ensemble consists of reactive trajectories and possesses all the information necessary for the understanding of the mechanism and dynamics of important condensed phase processes. However, quantitative description of the properties of the transition path ensemble is far from being established. Here, with numerical calculations on a model system, the equipartition terms defined in thermal equilibrium were for the first time estimated in the transition path ensemble. It was not surprising to observe that the energy was not equally distributed among all the coordinates. However, the energies distributed on a pair of conjugated coordinates remained equal. Higher energies were observed to be distributed on several coordinates, which are highly coupled to the reaction coordinate, while the rest were almost equally distributed. In addition, the ensemble-averaged energy on each coordinate as a function of time was also quantified. These quantitative analyses on energy distributions provided new insights into the transition path ensemble.

Ion dissolution mechanism and kinetics at kink sites on NaCl surfaces

Desolvation barriers are present for solute-solvent exchange events, such as ligand binding to an enzyme active site, during protein folding, and at battery electrodes. For solution-grown crystals, desolvation at kink sites can be the rate-limiting step for growth. However, desolvation and the associated kinetic barriers are poorly understood. In this work, we use rare-event simulation techniques to investigate attachment/detachment events at kink sites of a NaCl crystal in water. We elucidate the desolvation mechanism and present an optimized reaction coordinate, which involves one solute collective variable and one solvent collective variable. The attachment/detachment pathways for Na⁺ and Cl^- are qualitatively similar, with quantitative differences that we attribute to different ion sizes and solvent coordination. The attachment barriers primarily result from kink site desolvation, while detachment barriers largely result from breaking ion-crystal bonds. We compute ion detachment rates from kink sites and compare with results from an independent study. We anticipate that the reaction coordinate and desolvation mechanism identified in this work may be applicable to other alkali halides.

Identification of simple reaction coordinates from complex dynamics

Reaction coordinates are widely used throughout chemical physics to model and understand complex chemical transformations. We introduce a definition of the natural reaction coordinate, suitable for condensed phase and biomolecular systems, as a maximally predictive one-dimensional projection. We then show that this criterion is uniquely satisfied by a dominant eigenfunction of an integral operator associated with the ensemble dynamics. We present a new sparse estimator for these eigenfunctions which can search through a large candidate pool of structural order parameters and build simple, interpretable approximations that employ only a small number of these order parameters. Example applications with a small molecule's rotational dynamics and simulations of protein conformational change and folding show that this approach can filter through statistical noise to identify simple reaction coordinates from complex dynamics.

A benchmark for reaction coordinates in the transition path ensemble

The molecular mechanism of a reaction is embedded in its transition path ensemble, the complete collection of reactive trajectories. Utilizing the information in the transition path ensemble alone, we developed a novel metric, which we termed the emergent potential energy, for distinguishing reaction coordinates from the bath modes. The emergent potential energy can be understood as the average energy cost for making a displacement of a coordinate in the transition path ensemble. Where displacing a bath mode invokes essentially no cost, it costs significantly to move the reaction coordinate. Based on some general assumptions of the behaviors of reaction and bath coordinates in the transition path ensemble, we proved theoretically with statistical mechanics that the emergent potential energy could serve as a benchmark of reaction coordinates and demonstrated its effectiveness by applying it to a prototypical system of biomolecular dynamics. Using the emergent potential energy as guidance, we developed a committor-free and intuition-independent method for identifying reaction coordinates in complex systems. We expect this method to be applicable to a wide range of reaction processes in complex biomolecular systems.

Reducing the cost of evaluating the committor by a fitting procedure

Correct identification of reaction coordinates in complex systems is essential for understanding the mechanisms of their reaction dynamics. Existing methods for identifying reaction coordinates typically require knowledge of the committor--the probability of a given configuration to reach the product basin. The high computational cost of evaluating committors has limited applications of methods for identifying reaction coordinates. We proposed a fitting procedure that can reduce the cost of evaluating committors by an order of magnitude or more. The method only requires evaluating the committors of a few configurations in a transition path by the standard and costly shooting procedure. The committors of the other configurations are then estimated with great accuracy by a sigmoid function derived from fitting the few numerically evaluated committors. The method has been systematically tested on a model system of a Brownian particle moving in a one-dimensional double-well potential, and a small biomolecular system--the isomerization of alanine dipeptide in vacuum and in explicit water.

Understanding the rates and molecular mechanism of water-exchange around aqueous ions using molecular simulations

Solvation processes occurring around aqueous ions are of fundamental importance in physics, chemistry, and biology. Over the past few decades, several experimental and theoretical studies were devoted to understanding ion solvation and the processes involved in it. In this article, we present a summary of our recent efforts that, through computer simulations, focused on providing a comprehensive understanding of solvent-exchange processes around aqueous ions. To accomplish these activities, we have looked at the mechanistic properties associated with the water-exchange process, such as potentials of mean force, time-dependent transmission coefficients, and the corresponding rate constants using transition state theory, the reactive flux method, and Grote-Hynes treatments of the dynamic response of the solvent.

Recent developments in methods for identifying reaction coordinates

In the study of rare events in complex systems with many degrees of freedom, a key element is to identify the reaction coordinates of a given process. Over recent years, a number of methods and protocols have been developed to extract the reaction coordinates based on limited information from molecular dynamics simulations. In this review, we provide a brief survey over a number of major methods developed in the past decade, some of which are discussed in greater detail, to provide an overview of the problems that are partially solved and challenges that still remain. A particular emphasis has been placed on methods for identifying reaction coordinates that are related to the committor.