Developing an accurate interatomic potential model is a prerequisite for achieving reliable results from classical molecular dynamics (CMD) simulations; however, most of the potentials are biased as specific simulation purposes or conditions are considered in the parameterization. For developing an unbiased potential, a finite‐temperature dynamics machine learning (FTD‐ML) approach is proposed, and its processes and feasibility are demonstrated using the Buckingham potential model and aluminum (Al) as an example. Compared with conventional machine learning approaches, FTD‐ML exhibits three distinguished features: 1) FTD‐ML intrinsically incorporates more extensive configurational and conditional space for enhancing the transferability of developed potentials; 2) FTD‐ML employs various properties calculated directly from CMD, for ML model training and prediction validation against experimental data instead of first‐principles data; 3) FTD‐ML is much more computationally cost effective than first‐principles simulations, especially when the system size increases over 103atoms as employed in this research for ensuring reliable training data. The Al Buckingham potential developed by the FTD‐ML approach exhibits good performance for general simulation purposes. Thus, the FTD‐ML approach is expected to contribute to a fast development of interatomic potential model suitable for various simulation purposes and conditions, without limitation of model type, while maintaining experimental‐level accuracy.
This work presents Neural Equivariant Interatomic Potentials (NequIP), an E(3)-equivariant neural network approach for learning interatomic potentials from ab-initio calculations for molecular dynamics simulations. While most contemporary symmetry-aware models use invariant convolutions and only act on scalars, NequIP employs E(3)-equivariant convolutions for interactions of geometric tensors, resulting in a more information-rich and faithful representation of atomic environments. The method achieves state-of-the-art accuracy on a challenging and diverse set of molecules and materials while exhibiting remarkable data efficiency. NequIP outperforms existing models with up to three orders of magnitude fewer training data, challenging the widely held belief that deep neural networks require massive training sets. The high data efficiency of the method allows for the construction of accurate potentials using high-order quantum chemical level of theory as reference and enables high-fidelity molecular dynamics simulations over long time scales.
more » « less- NSF-PAR ID:
- 10381731
- Publisher / Repository:
- Nature Publishing Group
- Date Published:
- Journal Name:
- Nature Communications
- Volume:
- 13
- Issue:
- 1
- ISSN:
- 2041-1723
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
more » « less
Abstract -
Li, Jinyan (Ed.)more » « less
Artificial intelligence-powered protein structure prediction methods have led to a paradigm-shift in computational structural biology, yet contemporary approaches for predicting the interfacial residues (i.e., sites) of protein-protein interaction (PPI) still rely on experimental structures. Recent studies have demonstrated benefits of employing graph convolution for PPI site prediction, but ignore symmetries naturally occurring in 3-dimensional space and act only on experimental coordinates. Here we present EquiPPIS, an E(3) equivariant graph neural network approach for PPI site prediction. EquiPPIS employs symmetry-aware graph convolutions that transform equivariantly with translation, rotation, and reflection in 3D space, providing richer representations for molecular data compared to invariant convolutions. EquiPPIS substantially outperforms state-of-the-art approaches based on the same experimental input, and exhibits remarkable robustness by attaining better accuracy with predicted structural models from AlphaFold2 than what existing methods can achieve even with experimental structures. Freely available at
https://github.com/Bhattacharya-Lab/EquiPPIS , EquiPPIS enables accurate PPI site prediction at scale. -
Neural Network potentials are developed which accurately make and break bonds for use in molecular simulations. We report a neural network potential that can describe the potential energy surface for carbon–carbon bond dissociation with less than 1 kcal mol−1 error compared to complete active space second-order perturbation theory (CASPT2), and maintains this accuracy for both the minimum energy path and molecular dynamic calculations up to 2000 K. We utilize a transfer learning algorithm to develop neural network potentials to generate potential energy surfaces; this method aims to use the minimum amount of CASPT2 data on small systems to train neural network potentials while maintaining excellent transferability to larger systems. First, we generate homolytic carbon–carbon bond dissociation data of small size alkanes with density functional theory (DFT) energies to train the potentials to accurately predict bond dissociation at the DFT level. Then, using transfer learning, we retrained the neural network potential to the CASPT2 level of accuracy. We demonstrate that the neural network potential only requires bond dissociation data of a few small alkanes to accurately predict bond dissociation energy in larger alkanes. We then perform additional training on molecular dynamic simulations to refine our neural network potentials to obtain high accuracy for general use in molecular simulation. This training algorithm is generally applicable to any type of bond or any level of theory and will be useful for the generation of new reactive neural network potentials.more » « less
-
more » « less
Abstract A simultaneously accurate and computationally efficient parametrization of the potential energy surface of molecules and materials is a long-standing goal in the natural sciences. While atom-centered message passing neural networks (MPNNs) have shown remarkable accuracy, their information propagation has limited the accessible length-scales. Local methods, conversely, scale to large simulations but have suffered from inferior accuracy. This work introduces Allegro, a strictly local equivariant deep neural network interatomic potential architecture that simultaneously exhibits excellent accuracy and scalability. Allegro represents a many-body potential using iterated tensor products of learned equivariant representations without atom-centered message passing. Allegro obtains improvements over state-of-the-art methods on QM9 and revMD17. A single tensor product layer outperforms existing deep MPNNs and transformers on QM9. Furthermore, Allegro displays remarkable generalization to out-of-distribution data. Molecular simulations using Allegro recover structural and kinetic properties of an amorphous electrolyte in excellent agreement with ab-initio simulations. Finally, we demonstrate parallelization with a simulation of 100 million atoms.
-
The development of reliable, yet computationally efficient interatomic forcefields is key to facilitate the modeling of glasses. However, the parameterization of novel forcefields is challenging as the high number of parameters renders traditional optimization methods inefficient or subject to bias. Here, we present a new parameterization method based on machine learning, which combines ab initio molecular dynamics simulations and Bayesian optimization. By taking the example of glassy silica, we show that our method yields a new interatomic forcefield that offers an unprecedented agreement with ab initio simulations. This method offers a new route to efficiently parameterize new interatomic forcefields for disordered solids in a non-biased fashion.more » « less
