High-throughput Simulations Research Articles

Accelerating Molecular Dynamics Simulations Using Socket-Based Interprocess Communication.

Molecular dynamics (MD) simulations are essential for studying the time evolution of molecular systems. Still, their efficiency is often bottlenecked by file-based interprocess communication (IPC) between MD and electronic structure programs. We present a socket-based IPC implementation that dramatically accelerates MD simulations, reducing the computational time by >10-fold compared to those of traditional file-based methods. Our approach, applied to nonadiabatic molecular dynamics with the Newton-X program, eliminates disk read/write overhead, allowing for faster simulations over longer time scales. This method opens the door to more efficient high-throughput simulations, providing new opportunities for exploring complex molecular processes in real time.

StreaMD: the toolkit for high-throughput molecular dynamics simulations.

Molecular dynamics simulations serve as a prevalent approach for investigating the dynamic behaviour of proteins and protein-ligand complexes. Due to its versatility and speed, GROMACS stands out as a commonly utilized software platform for executing molecular dynamics simulations. However, its effective utilization requires substantial expertise in configuring, executing, and interpreting molecular dynamics trajectories. Existing automation tools are constrained in their capability to conduct simulations for large sets of compounds with minimal user intervention, or in their ability to distribute simulations across multiple servers. To address these challenges, we developed a Python-based tool that streamlines all phases of molecular dynamics simulations, encompassing preparation, execution, and analysis. This tool minimizes the required knowledge for users engaging in molecular dynamics simulations and can efficiently operate across multiple servers within a network or a cluster. Notably, the tool not only automates trajectory simulation but also facilitates the computation of free binding energies for protein-ligand complexes and generates interaction fingerprints across the trajectory. Our study demonstrated the applicability of this tool on several benchmark datasets. Additionally, we provided recommendations for end-users to effectively utilize the tool.Scientific contributionThe developed tool, StreaMD, is applicable to different systems (proteins, ligands and their complexes including co-factors) and requires a little user knowledge to setup and run molecular dynamics simulations. Other features of StreaMD are seamless integration with calculation of MM-GBSA/PBSA binding free energies and protein-ligand interaction fingerprints, and running of simulations within distributed environments. All these will facilitate routine and massive molecular dynamics simulations.

Machine learning assisted crystallographic reconstruction from atom probe tomographic images

Atom probe tomography (APT) is a powerful technique for three-dimensional (3D) atomic-scale imaging, enabling the accurate analysis on the compositional distribution at the nanoscale. How to accurately reconstruct crystallographic information from APT data, however, is still a great challenge due to the intrinsic nature of the APT technique. In this paper, we propose a novel approach that consists of the modified forward simulation process and the backward machine learning process to recover the tested crystal from APT data. The high-throughput forward simulations on Al single crystals of different orientations generate 10 000 original 3D images and data augmentation is implemented on the original images, resulting in 100 000 3D images. The big data allows the development of deep learning models and three deep learning algorithms of Convolutional Neural Network (CNN), Vision Transformer (ViT), and Variational Autoencoder (VAE) are used in the backward process. After training, the ViT model performs superior than the CNN and VAE models, which can recover the crystalline orientation outstandingly, as evaluated by the coefficient of determinationR2and the Mean Percent Error (MPE), viz.,R2= 0.93 and MPE = 0.43%,R2= 0.97 and MPE = 0.35%, andR2= 0.93 and MPE = 0.77% for the rotation anglesϕ,ψandθ, respectively, on the test dataset. The present work clearly demonstrates the capability of deep learning models in the recovery of the tested crystals from APT data, thereby paving the way for the further development of large artificial intelligent models of APT.

Optimized performance of membrane-based desalination by high-throughput molecular dynamic simulations and machine learning analysis

The influence of pore size and hydrophilicity on the permeance of reverse osmosis (RO) membranes has been mostly focused. However, their influence is hardly to be clearly identified as these two kinds of factor interfere with each other. In this work, high-throughput molecular dynamics (HTMD) simulations with CNTs are used to extensively produce the data of water permeance and NaCl rejection. These data are then analyzed by machine learning (ML) method to obtain the optimized desalination performance. The HTMD results indicate that the pressure drop has little effect on the water permeance. Moreover, rising pore size and degrading hydrophilicity will generally boost water permeance but will somehow sacrifice the NaCl rejection. The interference effect between pore size and hydrophilicity is also found in this work, the mechanism of which is then revealed from molecular level. Additionally, ML is applied to analyze the abundant data of water permeance and NaCl rejection. The optimal conditions are identified to achieve the highest water permeance with 100% NaCl rejection, which are also validated via additional MD simulations. This work suggests that the integration of HTMD and ML promises the future of designing new kind of RO membranes for better performance.

Electronic structure simulations in the cloud computing environment.

The transformative impact of modern computational paradigms and technologies, such as high-performance computing (HPC), quantum computing, and cloud computing, has opened up profound new opportunities for scientific simulations. Scalable computational chemistry is one beneficiary of this technological progress. The main focus of this paper is on the performance of various quantum chemical formulations, ranging from low-order methods to high-accuracy approaches, implemented in different computational chemistry packages and libraries, such as NWChem, NWChemEx, Scalable Predictive Methods for Excitations and Correlated Phenomena, ExaChem, and Fermi-Löwdin orbital self-interaction correction on Azure Quantum Elements, Microsoft's cloud services platform for scientific discovery. We pay particular attention to the intricate workflows for performing complex chemistry simulations, associated data curation, and mechanisms for accuracy assessment, which is demonstrated with the Arrows automated workflow for high throughput simulations. Finally, we provide a perspective on the role of cloud computing in supporting the mission of leadership computational facilities.

Extreme time extrapolation capabilities and thermodynamic consistency of physics-inspired neural networks for the 3D microstructure evolution of materials via Cahn–Hilliard flow

Abstract A Convolutional Recurrent Neural Network (CRNN) is trained to reproduce the evolution of the spinodal decomposition process in three dimensions as described by the Cahn–Hilliard equation. A specialized, physics-inspired architecture is proven to provide close accordance between the predicted evolutions and the ground truth ones obtained via conventional integration schemes. The method can accurately reproduce the evolution of microstructures not represented in the training set at a fraction of the computational costs. Extremely long-time extrapolation capabilities are achieved, up to reaching the theoretically expected equilibrium state of the system, consisting of a layered, phase-separated morphology, despite the training set containing only relatively-short, initial phases of the evolution. Quantitative accordance with the decay rate of the free energy is also demonstrated up to the late coarsening stages, proving that this class of machine learning approaches can become a new and powerful tool for the long timescale and high throughput simulation of materials, while retaining thermodynamic consistency and high-accuracy.

High-throughput molecular simulations of SARS-CoV-2 receptor binding domain mutants quantify correlations between dynamic fluctuations and protein expression.

Prediction of protein fitness from computational modeling is an area of active research in rational protein design. Here, we investigated whether protein fluctuations computed from molecular dynamics simulations can be used to predict the expression levels of SARS-CoV-2 receptor binding domain (RBD) mutants determined in the deep mutational scanning experiment of Starr et al. [Science (New York, N.Y.) 2022, 377, 420] Specifically, we performed more than 0.7 milliseconds of molecular dynamics (MD) simulations of 557 mutant RBDs in triplicate to achieve statistical significance under various simulation conditions. Our results show modest but significant anticorrelation in the range [-0.4, -0.3] between expression and RBD protein flexibility. A simple linear regression machine learning model achieved correlation coefficients in the range [0.7, 0.8], thus outperforming MD-based models, but required about 25 mutations at each residue position for training.

Modeling chemical bioaccumulation in snakes, part 2: Model testing and high-throughput simulation

In part 2 of the physiologically based kinetic (PBK) model for snakes, using default and generic input values, the simulation outcomes showed that chemicals with moderate lipophilicity, low volatility, and low biotransformability exhibited a high potential for biotransfer in the snake’s skin. Furthermore, the inclusion or exclusion of the skin compartment in the PBK model had a substantial impact on the fate, transport, and distribution of these chemicals within the snake’s body. In comparison to the elimination routes via blood transport and volatilization, the shedding of skin and growth processes did not contribute substantially to the overall kinetics of chemical elimination from the skin for most chemicals. The proposed model has demonstrated a consistent alignment with the observed patterns of chemical distribution, as supported by certain experimental data. Furthermore, it has the potential to provide an explanation for and enhance the understanding of the discrepancies found in other field observations. The modeling exercise clearly illustrated the significant role of the skin compartment in the biotransfer of chemicals within the snake’s body and highlighted the importance of including the snake’s physiological features into the PBK model. To further enhance the model’s performance and accuracy, additional research focused on obtaining specific physiological data pertaining to snakes would be beneficial.

Ultrafast manipulations of nanoscale skyrmioniums

The advancement of next-generation magnetic devices depends on fast manipulating magnetic microstructures at the nanoscale. A universal method is presented for rapid and reliable generating, controlling, and driving nano-scale skyrmioniums, through high-throughput micromagnetic simulations. Ultrafast switches are realized between skyrmionium and skyrmion states and rapidly change their polarities in monolayer magnetic nanodisks by perpendicular magnetic fields. The transition mechanism by alternating magnetic fields differs from that under steady magnetic fields. New skyrmionic textures, such as flower-like and windmill-like skyrmions, are discovered. Moreover, this nanoscale skyrmionium can move rapidly and stably in nanoribbons using weaker spin-polarized currents. Explicit discussions are held regarding the physical mechanisms involved in ultrafast manipulations of skyrmioniums. This work provides further physical insights into the manipulation and application of topological skyrmionic structures for developing low-power consumption and nanostorage devices.

Chicoric acid acts as an ALOX15 inhibitor to prevent ferroptosis in asthma

BackgroundChicoric acid (CA) is a crucial immunologically active compound found in chicory and echinacea, possessing a range of biological activities. Ferroptosis, a type of iron-dependent cell death induced by lipid peroxidation, plays a key role in the development and advancement of asthma. Targeting ferroptosis could be a potential therapeutic strategy for treating asthma. PurposeThe purpose of this study was to explore the screening of ALOX15, a pivotal target of ferroptosis in asthma, and potential therapeutic agents, as well as to investigate the promising potential of CA as an ALOX15 inhibitor for modulating ferroptosis in asthma. MethodsThrough high-throughput data processing of bronchial epithelial RNA from asthma patients using bioinformatics and machine learning, the key target of ferroptosis in asthma, ALOX15, was identified. An inhibitor of ALOX15 was then obtained through high-throughput molecular docking and molecular dynamics simulation tests. In vitro experiments were conducted using a 16HBE cell model induced by house dust mite (HDM) and lipopolysaccharide (LPS), which were treated with the ALOX15 inhibitor (PD146176), CA treatment, or ALOX15 knockdown. In vivo experiments were also carried out using a mouse model induced by HDM and LPS. ResultsThe composite model of ALOX15 and CA in molecular dynamics simulations shows good stability and flexibility. Network pharmacological analysis reveals that CA regulates ferroptosis through ALOX15 in treating asthma. In vitro studies show that ALOX15 is highly expressed in HDM and LPS treatments, while CA inhibits HDM and LPS-induced ferroptosis in 16HBE cells by reducing ALOX15 expression. Knockdown of ALOX15 has the opposite effect. Metabolomics analysis identifies key compounds associated with ferroptosis, including L-Targinine, eicosapentaenoic acid, 16-hydroxy hexadecanoic acid, and succinic acid. In vivo experiments demonstrate that CA suppresses ALOX15 expression, inhibits ferroptosis, and improves asthma symptoms in mice. ConclusionOur research initially identified CA as a promising asthma treatment that effectively blocks ferroptosis by specifically targeting ALOX15. This study not only highlights CA as a potential therapeutic agent for asthma but also introduces novel targets and treatment options for this condition, along with innovative approaches for utilizing natural compounds to target diseases associated with ferroptosis.

Data-driven modeling of dislocation mobility from atomistics using physics-informed machine learning

Dislocation mobility, which dictates the response of dislocations to an applied stress, is a fundamental property of crystalline materials that governs the evolution of plastic deformation. Traditional approaches for deriving mobility laws rely on phenomenological models of the underlying physics, whose free parameters are in turn fitted to a small number of intuition-driven atomic scale simulations under varying conditions of temperature and stress. This tedious and time-consuming approach becomes particularly cumbersome for materials with complex dependencies on stress, temperature, and local environment, such as body-centered cubic crystals (BCC) metals and alloys. In this paper, we present a novel, uncertainty quantification-driven active learning paradigm for learning dislocation mobility laws from automated high-throughput large-scale molecular dynamics simulations, using Graph Neural Networks (GNN) with a physics-informed architecture. We demonstrate that this Physics-informed Graph Neural Network (PI-GNN) framework captures the underlying physics more accurately compared to existing phenomenological mobility laws in BCC metals.

Establishing the longitudinal hemodynamic mapping framework for wearable-driven coronary digital twins

Understanding the evolving nature of coronary hemodynamics is crucial for early disease detection and monitoring progression. We require digital twins that mimic a patient’s circulatory system by integrating continuous physiological data and computing hemodynamic patterns over months. Current models match clinical flow measurements but are limited to single heartbeats. To this end, we introduced the longitudinal hemodynamic mapping framework (LHMF), designed to tackle critical challenges: (1) computational intractability of explicit methods; (2) boundary conditions reflecting varying activity states; and (3) accessible computing resources for clinical translation. We show negligible error (0.0002–0.004%) between LHMF and explicit data of 750 heartbeats. We deployed LHMF across traditional and cloud-based platforms, demonstrating high-throughput simulations on heterogeneous systems. Additionally, we established LHMFC, where hemodynamically similar heartbeats are clustered to avoid redundant simulations, accurately reconstructing longitudinal hemodynamic maps (LHMs). This study captured 3D hemodynamics over 4.5 million heartbeats, paving the way for cardiovascular digital twins.

Phase-field modeling of crack growth under coupled creep-fatigue

Crack growth under coupled creep-fatigue strongly influences service life of metallic components operating at high temperatures. However, there lack numerical models that enable direct simulations of crack growth and plasticity development under coupled creep-fatigue. In this study, a unified phase-field model is developed for simulating crack growth under fatigue, creep, and coupled creep-fatigue. This model is able to reproduce the Paris laws for both brittle and ductile materials in pure fatigue mode, which shows that the degradation rate due to cyclic stress in ductile materials is generally larger than that in brittle materials. Fatigue simulations with different yield strengths demonstrate that plasticity facilitates crack growth in typical ductile materials. High-throughput simulations are performed under coupled creep-fatigue conditions, with varying hold-stress levels and hold-time. By analyzing crack growth rate under different hold-time, a creep-fatigue interaction term is obtained. Simulation results demonstrate that the creep-fatigue interaction is caused by smaller stress gradient and enhanced degradation rate ahead of the crack tips under coupled creep-fatigue, compared to the situation of pure fatigue. This work reveals the origin of different Paris-law exponents in brittle and ductile materials and demonstrates how the interplay between creep and fatigue affects crack growth.

Gas permeability, diffusivity, and solubility in polymers: Simulation-experiment data fusion and multi-task machine learning

Machine learning (ML) models for predicting gas permeability through polymers have traditionally relied on experimental data. While these models exhibit robustness within familiar chemical domains, reliability wanes when applied to new spaces. To address this challenge, we present a multi-tiered multi-task learning framework empowered with advanced machine-crafted polymer fingerprinting algorithms and data fusion techniques. This framework combines scarce “high-fidelity” experimental data with abundant diverse “low-fidelity” simulation or synthetic data, resulting in predictive models that display a high level of generalizability across novel chemical spaces. Additionally, this multi-task scheme capitalizes on known physics and interrelated properties, such as gas diffusivity and solubility, both of which are closely tied to permeability. By amalgamating high throughput generated simulation data with available experimental data for gas permeability, diffusivity, and solubility for various gases, we construct multi-task deep learning models. These models can simultaneously predict all three properties for all gases under consideration, with markedly enhanced predictive accuracy, particularly compared to traditional models reliant solely on experimental data for a singular property. This strategy underscores the potential of coupling high-throughput classical simulations with data fusion methodologies to yield state-of-the-art property predictors, especially when experimental data for targeted properties is scarce.

Unveiling the quantitative relationship between microstructural features and quasi-static tensile properties in dual-phase titanium alloys based on data-driven neural networks

The quasi-static mechanical properties of α+β dual-phase titanium alloys are susceptible to their microstructural features, presenting a complex, high-dimensional nonlinear relationship, which hinders the rapid development of high-performance materials. In this work, 4065 micro-representative models were virtually constructed with varying volume fractions of α and β phases and characteristic dimensions via high-throughput finite element simulation, incorporating a cohesive zone model to simulate the interfaces between the two phases. Especially, the established representative models were experimentally verified by two groups of real material microstructures, and the results showed that the relative errors were not more than 9.5 % in microstructural characteristics and quasi-static mechanical properties. Afterward, a neural network model was developed to correlate the quasi-static tensile properties with the microstructural features of the dual-phase TC6 titanium alloys, achieving an 88.2 % accuracy in predicting overall mechanical performance. Utilizing the Shaply Additive Explanation method, it was found that the primary α phase's volume fraction and the secondary α phase's width were the most significant microstructural features affecting quasi-static strength. Specifically, the volume fraction of the primary α phase and the width of the secondary α phase negatively affected strength, while the width of the secondary α phase positively influenced plasticity. Notably, the primary α phase's volume fraction had a quadratic curve pattern of influence on plasticity. The intrinsic mechanisms behind these laws were further revealed based on local stress-strain responses and crack propagation analysis. Ultimately, the optimal microstructural features with strength-plasticity balance were identified through the lower threshold method: a secondary α phase width of about 1 μm and a primary α phase volume fraction ranging from 0.1 to 0.2, effectively facilitating microstructure design.

Combining machine learning and molecular simulation to explore MOF materials that contribute to CF4/N2 separation

High-throughput molecular simulations and machine learning algorithms have been widely used to identify promising metal–organic frameworks for gas separation. However, most studies are limited to screening high-performance materials from existing databases, which fails to fully utilize the predictive function of machine learning. This paper combines genetic algorithms and high-throughput screening to deeply mine anion-pillared MOF (APMOF) performance feature relationships and predict high-performance materials that are not in the database. Considering the actual industrial conditions (CF4/N2 = 10/90), we chose the ratio of CF4:N2 = 1:9 to simulate the adsorption separation of gas mixtures of MOF materials at a temperature of 298 K and a pressure of 1 bar. First, the CF4/N2 separation properties of MOFs in the APMOF library were obtained based on molecular simulations. Then, the filtered data were coded according to the method of “building block classification − structural interval categorization”. Then, the machine learning algorithm was used for model training to obtain a high-precision model. Finally, the tangent adaptive genetic algorithm was used to recombine the genes of the materials, and the new MOF materials were successfully reverse-engineered. The study found that the pore-limit diameter of APMOFs is most conducive to the separation of CF4/N2 by MOFs when the pore-limit diameter of APMOFs is within two times the molecular dynamics diameter of CF4. 134 MOF materials were predicted to have CF4/N2 selectivities exceeding 46.30. The use of organic ligands such as 4,4′-bipyridyl or 1,4-bis(4-pyridyl)benzene (bpb) increases the likelihood of these materials being high-performance for CF4/N2 separation. The combination of computational screening methods and machine learning can expedite the design and development of new high-performance MOFs. This approach can also provide guidance for the synthetic direction and pattern of materials, which can assist experimentalists in their synthesis.

High-throughput phase field simulation and machine learning for predicting the breakdown performance of all-organic composites

All-organic dielectric polymers are materials of choice for modern power electronics and high-density energy storage, and their performance can be significantly improved by doping trace amounts of organic molecular semiconductors with strong electron-affinity energy to suppress charge conduction losses. Insight into the breakdown mechanism of polymers/organic molecular semiconductor composites is essential for the design of high-performance dielectric polymers. This study investigates the impact of the doping concentration of organic molecular semiconductors, dielectric constants, and trap depths on the breakdown performance of dielectric polymers under high temperature and electric fields. A modified phase-field model, incorporating deep traps and carriers’ coulomb capture radius, has been developed to facilitate high-throughput simulations of electrical breakdown in polymer/organic molecular semiconductor composites. This work accurately predicted the breakdown strength of all-organic composites using high-throughput phase-field simulation data as input for machine learning, which provides crucial theoretical support for designing all-organic composite dielectric polymers for energy storage capacitors under extreme conditions.

Frictional Properties of Two-Dimensional Materials: Data-Driven Machine Learning Predictive Modeling.

Friction, typically associated with reduced efficiency and reliability of machines and devices, occurs when two objects are displaced against each other. This is a strongly material-dependent phenomenon, and the emergence of many 2D materials has opened up new opportunities to design systems with desired tribological properties. Here, we combine high throughput simulations and machine learning models to develop a statistical approach of adhesion, van der Waals, and corrugation energies of a large dataset of monolayered materials. The machine learning models are used to predict these closely related to friction energetic properties and link them to easily accessible atomistic and monolayer features. This approach elevates the materials' perspective of frictional properties. It demonstrates that data-driven models are extremely useful in discovering important structure-property functionalities for frictional property interpretations as a fruitful route toward desired tribological materials.

Benchmarking of machine learning interatomic potentials for reactive hydrogen dynamics at metal surfaces

Simulations of chemical reaction probabilities in gas surface dynamics require the calculation of ensemble averages over many tens of thousands of reaction events to predict dynamical observables that can be compared to experiments. At the same time, the energy landscapes need to be accurately mapped, as small errors in barriers can lead to large deviations in reaction probabilities. This brings a particularly interesting challenge for machine learning interatomic potentials, which are becoming well-established tools to accelerate molecular dynamics simulations. We compare state-of-the-art machine learning interatomic potentials with a particular focus on their inference performance on CPUs and suitability for high throughput simulation of reactive chemistry at surfaces. The considered models include polarizable atom interaction neural networks (PaiNN), recursively embedded atom neural networks (REANN), the MACE equivariant graph neural network, and atomic cluster expansion potentials (ACE). The models are applied to a dataset on reactive molecular hydrogen scattering on low-index surface facets of copper. All models are assessed for their accuracy, time-to-solution, and ability to simulate reactive sticking probabilities as a function of the rovibrational initial state and kinetic incidence energy of the molecule. REANN and MACE models provide the best balance between accuracy and time-to-solution and can be considered the current state-of-the-art in gas-surface dynamics. PaiNN models require many features for the best accuracy, which causes significant losses in computational efficiency. ACE models provide the fastest time-to-solution, however, models trained on the existing dataset were not able to achieve sufficiently accurate predictions in all cases.

Incorporating Prior Knowledge in the Seeds of Adaptive Sampling Molecular Dynamics Simulations of Ligand Transport in Enzymes with Buried Active Sites.

Because most proteins have buried active sites, protein tunnels or channels play a crucial role in the transport of small molecules into buried cavities for enzymatic catalysis. Tunnels can critically modulate the biological process of protein-ligand recognition. Various molecular dynamics methods have been developed for exploring and exploiting the protein-ligand conformational space to extract high-resolution details of the binding processes, a recent example being energetically unbiased high-throughput adaptive sampling simulations. The current study systematically contrasted the role of integrating prior knowledge while generating useful initial protein-ligand configurations, called seeds, for these simulations. Using a nontrivial system of a haloalkane dehalogenase mutant with multiple transport tunnels leading to a deeply buried active site, simulations were employed to derive kinetic models describing the process of association and dissociation of the substrate molecule. The most knowledge-based seed generation enabled high-throughput simulations that could more consistently capture the entire transport process, explore the complex network of transport tunnels, and predict equilibrium dissociation constants, koff/kon, on the same order of magnitude as experimental measurements. Overall, the infusion of more knowledge into the initial seeds of adaptive sampling simulations could render analyses of transport mechanisms in enzymes more consistent even for very complex biomolecular systems, thereby promoting drug development efforts and the rational design of enzymes with buried active sites.