SO3LR Tauri: Generally Applicable Force Fields for Reactive (Bio)Molecular Simulations Covering the Periodic Table
Abstract
Machine-learned force fields (MLFFs) aim to create a unified model that describes all major classes of (bio)molecules at quantum-mechanical accuracy. We present SO3LR Tauri, a family of MLFFs developed toward this goal, coupling a fast SO3krates network to explicit, physics-based interaction modules: pair-specific short-range repulsion, environment-dependent dispersion, and electrostatics, each improved beyond its original formulation. The models are trained on 148 million structures spanning 85 elements, drawn from several quantum-chemical datasets. Against coupled-cluster references, the models accurately reproduce nucleobase-pair and charged protein-ligand interactions, carbohydrate conformer ordering, and molecular dipoles. They also resolve the transition pathways of organic and enzymatic reactions. Dynamical simulations recover experimental densities of water and organic solvents, the hydration structure of mono- and divalent ions, the pKa values of amino-acid side chains, and the structure and dynamics of proteins, membranes, nucleic acids, glycoproteins, and metal-containing biomolecules, all without system-specific fitting. Released in three sizes, the models reach 0.7-3.2 μs/atom and scale to 27-84 million atoms on a single GPU. Much work remains to be done: achieving predictive accuracy for static and dynamic liquid properties, reaction barriers, coverage of chemical and conformational space, and the sampling and inference speed needed to describe biomolecular processes.
Related
- Reactive molecular dynamics simulations of atenolol first steps degradation by 9CL6 ammonia monooxygenase
- Interpretable Machine Learning of Hydrogen Production from Reactive Molecular Dynamics Simulations of Polyethylene Oxidative Gasification
- Reactive Molecular Dynamics Simulations of Mechanochemical Ammonia Production
- Interfacial Mechanisms of Reactive Dye Adsorption on a Waste-Cotton Derived PVA/TiO2/Cellulose Biofilm: A Molecular Dynamics and Physics-Informed Predictive Framework
- Crystallinity-Regulated Hydration, Molecular Mobility, and Mechanical Softening of Poly(ether ether ketone): A Reactive Molecular Dynamics Study
BibTeX
@misc{kabylda2026so3lr,
title = {{SO3LR Tauri: Generally Applicable Force Fields for Reactive (Bio)Molecular Simulations Covering the Periodic Table}},
author = {Adil Kabylda and Florian N. Brünig and Miguel Gallegos and J. Thorben Frank and Evgeny Moerman and Oliver T. Unke and Stefan Chmiela and Klaus-Robert Müller and Alexandre Tkatchenko},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.06237},
eprint = {2611.06237},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.06237},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - SO3LR Tauri: Generally Applicable Force Fields for Reactive (Bio)Molecular Simulations Covering the Periodic Table
AU - Adil Kabylda
AU - Florian N. Brünig
AU - Miguel Gallegos
AU - J. Thorben Frank
AU - Evgeny Moerman
AU - Oliver T. Unke
AU - Stefan Chmiela
AU - Klaus-Robert Müller
AU - Alexandre Tkatchenko
PY - 2026
DA - 2026/11/04
PB - arXiv
DO - 10.48550/arxiv.2611.06237
UR - https://arxiv.org/abs/2611.06237
AB - Machine-learned force fields (MLFFs) aim to create a unified model that describes all major classes of (bio)molecules at quantum-mechanical accuracy. We present SO3LR Tauri, a family of MLFFs developed toward this goal, coupling a fast SO3krates network to explicit, physics-based interaction modules: pair-specific short-range repulsion, environment-dependent dispersion, and electrostatics, each improved beyond its original formulation. The models are trained on 148 million structures spanning 85 elements, drawn from several quantum-chemical datasets. Against coupled-cluster references, the models accurately reproduce nucleobase-pair and charged protein-ligand interactions, carbohydrate conformer ordering, and molecular dipoles. They also resolve the transition pathways of organic and enzymatic reactions. Dynamical simulations recover experimental densities of water and organic solvents, the hydration structure of mono- and divalent ions, the pKa values of amino-acid side chains, and the structure and dynamics of proteins, membranes, nucleic acids, glycoproteins, and metal-containing biomolecules, all without system-specific fitting. Released in three sizes, the models reach 0.7-3.2 μs/atom and scale to 27-84 million atoms on a single GPU. Much work remains to be done: achieving predictive accuracy for static and dynamic liquid properties, reaction barriers, coverage of chemical and conformational space, and the sampling and inference speed needed to describe biomolecular processes.
ER -