cs.LGNov 16, 2026

tmol: a GPU-accelerated implementation of the Rosetta energy function for machine learning

Authors: Andrew Leaver-Fay, Kieran Didi, Jeff Flatten, Alex Ford, Guangfeng Zhou, Nathaniel Corley, Hahnbeom Park, Maxwell W. Kazman, 

Organizations: Prescient Design, Genentech, South San Francisco, CA, USA. · Institute for Protein Design, University of Washington, Seattle, WA, USA. · NVIDIA Corp., Santa Clara, USA. · Department of Computer Science, University of Oxford, Oxford, UK. · 

Abstract

Physics-inspired forcefields complement generative machine learning models, showing better generalization beyond the training distribution. However, these forcefields are typically slow to evaluate, and their derivatives are difficult to integrate into ML frameworks. We introduce tmol, a GPU-native PyTorch library implementing the state-of-the-art RosettaGenFF forcefield. tmol supports efficient batched differentiable scoring and whole-structure refinement with up to 362-fold throughput increase (batched). Used as a reward in generative protein design, tmol increases hydrogen bonds and water bridges at designed interfaces.

Figures & tables

Explore similar work

CardsList
  1. Implicit Machine Learning Force Fields Accelerate Molecular Dynamics Simulations

    Jul 31, 2026Johannes Maeß, Leon Werner, J. Thorben Frank +5Implicit Neural RepresentationsMachine Learning Interatomic Potentials

  2. Generative Pseudo-Force Fields for Molecular Generation

    May 18, 2026Stefaan Simon Pierre Hessmann, Khaled Kahouli, Stefan Gugler +4Machine Learning Interatomic PotentialsDiffusion Model Sampling

  3. Generating Physically Consistent Molecules with Energy-Based Models

    May 18, 2026Christoph Griesbacher, Lea Bogensperger, Andreas Habring +1Energy-Based Models3D Molecular Generation