26177

NNP/MM: Fast molecular dynamics simulations with machine learning potentials and molecular mechanics

Raimondas Galvelis, Alejandro Varela-Rial, Stefan Doerr, Roberto Fino, Peter Eastman, Thomas E. Markland, John D. Chodera, Gianni De Fabritiis
Acellera Labs, C/ Doctor Trueta 183, 08005 Barcelona, Spain
arXiv:2201.08110 [q-bio.BM], (20 Jan 2022)

@misc{galvelis2022nnpmm,

   title={NNP/MM: Fast molecular dynamics simulations with machine learning potentials and molecular mechanics},

   author={Raimondas Galvelis and Alejandro Varela-Rial and Stefan Doerr and Roberto Fino and Peter Eastman and Thomas E. Markland and John D. Chodera and Gianni De Fabritiis},

   year={2022},

   eprint={2201.08110},

   archivePrefix={arXiv},

   primaryClass={q-bio.BM}

}

Parametric and non-parametric machine learning potentials have emerged recently as a way to improve the accuracy of bio-molecular simulations. Here, we present NNP/MM, an hybrid method integrating neural network potentials (NNPs) and molecular mechanics (MM). It allows to simulate a part of molecular system with NNP, while the rest is simulated with MM for efficiency. The method is currently available in ACEMD using OpenMM plugins to optimize the performance of NNPs. The achieved performance is slower but comparable to the state-of-the-art GPU-accelerated MM simulations. We validated NNP/MM by performing MD simulations of four protein-ligand complexes, where NNP is used for the intra-molecular interactions of a lignad and MM for the rest interactions. This shows that NNP can already replace MM for small molecules in protein-ligand simulations. The combined sampling of each complex is 1 microsecond, which are the longest simulations of NNP/MM ever reported. Finally, we have made the setup of the NNP/MM simulations simple and user-friendly.
No votes yet.
Please wait...

* * *

* * *

HGPU group © 2010-2024 hgpu.org

All rights belong to the respective authors

Contact us: