TorchMD: A Deep Learning Framework for Molecular Simulations

• DOI: 10.1021/acs.jctc.0c01343
• Publication Date: 2021-03-17T22:39:52Z
• Authors: Stefan Doerr; Maciej Majewski; Adrià Pérez; Andreas Krämer; Cecilia Clementi; Frank Noe; Toni Giorgino; Gianni De Fabritiis
• Keywords: Deep Learning Framework; FOS: Computer and information sciences; 0301 basic medicine; PyTorch arrays; Chemical Sciences not elsewhere classified; network potentials; Computer Science - Artificial Intelligence; Information Systems not elsewhere classified; Biophysics; FOS: Physical sciences; 01 natural sciences; 03 medical and health sciences; Molecular Simulations Molecular dyn.; Physics - Chemical Physics; simulating; 0103 physical sciences; Genetics; data-driven models; Amber all-atom simulations; Chemical Physics (physics.chem-ph); ab initio; Coulomb interactions; coarse-grained model; Computational Biology; Cell Biology; artificial intelligence; computational chemistry; force computations; machine learning; Artificial Intelligence (cs.AI); TorchMD; Neuroscience; Biological Sciences not elsewhere classified

Molecular dynamics simulations provide a mechanistic description of molecules by relying on empirical potentials. The quality and transferability such potentials can be improved leveraging data-driven models derived with machine learning approaches. Here, we present TorchMD, framework for molecular mixed classical All force computations including bond, angle, dihedral, Lennard-Jones, Coulomb interactions are expressed as PyTorch arrays operations. Moreover, TorchMD enables simulating neural network We validate it using standard Amber all-atom simulations, an ab initio potential, performing end-to-end training, finally coarse-grained model protein folding. believe that provides useful tool set to support Code data freely available at github.com/torchmd.