Training atomic neural networks using fragment-based data generated in virtual reality

Free energy along drug-protein binding pathways interactively sampled in virtual reality

We describe a two-step approach for combining interactive molecular dynamics in virtual reality (iMD-VR) with free energy (FE) calculation to explore the dynamics of biological processes at the molecular level. We refer to this combined approach as iMD-VR-FE. Stage one involves using a state-of-the-art 'human-in-the-loop' iMD-VR framework to generate a diverse range of protein-ligand unbinding pathways, benefitting from the sophistication of human spatial and chemical intuition. Stage two involves using the iMD-VR-sampled pathways as initial guesses for defining a path-based reaction coordinate from which we can obtain a corresponding free energy profile using FE methods. To investigate the performance of the method, we apply iMD-VR-FE to investigate the unbinding of a benzamidine ligand from a trypsin protein. The binding free energy calculated using iMD-VR-FE is similar for each pathway, indicating internal consistency. Moreover, the resulting free energy profiles can distinguish energetic differences between pathways corresponding to various protein-ligand conformations (e.g., helping to identify pathways that are more favourable) and enable identification of metastable states along the pathways. The two-step iMD-VR-FE approach offers an intuitive way for researchers to test hypotheses for candidate pathways in biomolecular systems, quickly obtaining both qualitative and quantitative insight.

Neural network potentials for chemistry: concepts, applications and prospects

Artificial Neural Networks (NN) are already heavily involved in methods and applications for frequent tasks in the field of computational chemistry such as representation of potential energy surfaces (PES) and spectroscopic predictions. This perspective provides an overview of the foundations of neural network-based full-dimensional potential energy surfaces, their architectures, underlying concepts, their representation and applications to chemical systems. Methods for data generation and training procedures for PES construction are discussed and means for error assessment and refinement through transfer learning are presented. A selection of recent results illustrates the latest improvements regarding accuracy of PES representations and system size limitations in dynamics simulations, but also NN application enabling direct prediction of physical results without dynamics simulations. The aim is to provide an overview for the current state-of-the-art NN approaches in computational chemistry and also to point out the current challenges in enhancing reliability and applicability of NN methods on a larger scale.

Exploring Complex Reaction Networks Using Neural Network-Based Molecular Dynamics Simulation

Ab initio molecular dynamics (AIMD) is an established method for revealing the reactive dynamics of complex systems. However, the high computational cost of AIMD restricts the explorable length and time scales. Here, we develop a fundamentally different approach using molecular dynamics simulations powered by a neural network potential to investigate complex reaction networks. This potential is trained via a workflow combining AIMD and interactive molecular dynamics in virtual reality to accelerate the sampling of rare reactive processes. A panoramic visualization of the complex reaction networks for decomposition of a novel high explosive (ICM-102) is achieved without any predefined reaction coordinates. The study leads to the discovery of new pathways that would be difficult to uncover if established methods were employed. These results highlight the power of neural network-based molecular dynamics simulations in exploring complex reaction mechanisms under extreme conditions at the ab initio level, pushing the limit of theoretical and computational chemistry toward the realism and fidelity of experiments.

Community Reaction Network Reduction for Constructing a Coarse-Grained Representation of Combustion Reaction Mechanisms

A community-reaction network reduction (CNR) approach is presented for mechanism reduction on the basis of a network-based community detection technique, a concept related to pre-equilibrium in chemical kinetics. In this method, the detailed combustion mechanism is first transformed into a weighted network, in which communities of species that have dense inner connections under the critical ignition conditions are identified. By analyzing the community partitions in different regions, we determine the effective functional groups and driving processes. Then, a skeletal model for the overall mechanism is deduced according to the network centrality data, including transition pathway identification and reaction-path flux. The CNR method is illustrated on the hydrogen autoignition system which has been extensively investigated, and a new reduced mechanism involving seven processes is proposed. Dynamics simulations employing the present CNR model show that the computed ignition time and distribution of major species on a wide range of temperature and pressure conditions are in accord with the experiments and results from other methods.

CARNOT: a Fragment-Based Direct Molecular Dynamics and Virtual-Reality Simulation Package for Reactive Systems

Traditionally, the study of reaction mechanisms of complex reaction systems such as combustion has been performed on an individual basis by optimizations of transition structure and minimum energy path or by reaction dynamics trajectory calculations for one elementary reaction at a time. It is effective, but time-consuming, whereas important and unexpected processes could have been missed. In this article, we present a direct molecular dynamics (DMD) approach and a virtual-reality simulation program, CARNOT, in which plausible chemical reactions are simulated simultaneously at finite temperature and pressure conditions. A key concept of the present ab initio molecular dynamics method is to partition a large, chemically reactive system into molecular fragments that can be adjusted on the fly of a DMD simulation. The theory represents an extension of the explicit polarization method to reactive events, called ReX-Pol. We propose a highest-and-lowest adapted-spin approximation to define the local spins of individual fragments, rather than treating the entire system by a delocalized wave function. Consequently, the present ab initio DMD can be applied to reactive systems consisting of an arbitrarily varying number of closed and open-shell fragments such as free radicals, zwitterions, and separate ions found in combustion and other reactions. A graph-data structure algorithm was incorporated in CARNOT for the analysis of reaction networks, suitable for reaction mechanism reduction. Employing the PW91 density functional theory and the 6-31+G(d) basis set, the capabilities of the CARNOT program were illustrated by a combustion reaction, consisting of 28 650 atoms, and by reaction network analysis that revealed a range of mechanistic and dynamical events. The method may be useful for applications to other types of complex reactions.

Human Action Recognition in Smart Cultural Tourism Based on Fusion Techniques of Virtual Reality and SOM Neural Network

Smart cultural tourism is the development trend of the future tourism industry. Virtual reality is an important tool to realize smart tourism. The reality of virtual reality mainly comes from human-computer interaction, which is closely related to human action recognition technology. Therefore, the research takes human action recognition as the research direction, uses a self-organizing mapping network (SOM) neural network to extract the key frame of action video, combines it with multi-feature vector method to recognize human action, and compares the recognition rate and user satisfaction of different recognition methods. The results show that the recognition rate of multi-feature voting human action recognition algorithm based on SOM neural network is 93.68% on UT-Kinect action, 59.06% on MSRDailyActivity3D, and the overall action recognition time is only 3.59 s. Within six months, the total profit of human-computer interactive virtual reality tourism project with SOM neural network multi-eigenvector as the core algorithm reached 422,000 yuan, and 88% of users expressed satisfaction after use. It shows that the proposed method has a good recognition rate and can give users effective feedback in time. It is hoped that this research has a certain reference value in promoting the development of human motion recognition technology.

Exploring human-guided strategies for reaction network exploration: Interactive molecular dynamics in virtual reality as a tool for citizen scientists

The emerging fields of citizen science and gamification reformulate scientific problems as games or puzzles to be solved. Through engaging the wider non-scientific community, significant breakthroughs may be made by analyzing citizen-gathered data. In parallel, recent advances in virtual reality (VR) technology are increasingly being used within a scientific context and the burgeoning field of interactive molecular dynamics in VR (iMD-VR) allows users to interact with dynamical chemistry simulations in real time. Here, we demonstrate the utility of iMD-VR as a medium for gamification of chemistry research tasks. An iMD-VR "game" was designed to encourage users to explore the reactivity of a particular chemical system, and a cohort of 18 participants was recruited to playtest this game as part of a user study. The reaction game encouraged users to experiment with making chemical reactions between a propyne molecule and an OH radical, and "molecular snapshots" from each game session were then compiled and used to map out reaction pathways. The reaction network generated by users was compared to existing literature networks demonstrating that users in VR capture almost all the important reaction pathways. Further comparisons between humans and an algorithmic method for guiding molecular dynamics show that through using citizen science to explore these kinds of chemical problems, new approaches and strategies start to emerge.

Machine learning meets chemical physics

Over recent years, the use of statistical learning techniques applied to chemical problems has gained substantial momentum. This is particularly apparent in the realm of physical chemistry, where the balance between empiricism and physics-based theory has traditionally been rather in favor of the latter. In this guest Editorial for the special topic issue on "Machine Learning Meets Chemical Physics," a brief rationale is provided, followed by an overview of the topics covered. We conclude by making some general remarks.