Preprint · 2026 · Open access

Fast training of bespoke SMIRNOFF-format molecular mechanics force fields using machine learning potentials

Finlay Clark, Thomas J. Pope, Sarah Maier, Simon Boothroyd, Josh T. Horton, Kevin Ryczko, Andrea Bortolato, Daniel J. Cole

ChemRxiv · Published 8 June 2026 · DOI 10.26434/chemrxiv.15004169/v2

Free PDFView at the publisherCite this paper

In brief

“Fast training of bespoke SMIRNOFF-format molecular mechanics force fields using machine learning potentials” is a 2026 preprint by Finlay Clark and 7 co-authors, published in ChemRxiv. OpenAlex counts 0 citations as of 16 November 2026. It is open access under a CC BY licence: the full text is free to read at ChemRxiv.

Citations
0
References
19
Authors
8

Abstract

Molecular mechanics force fields enable the thorough sampling required for applications such as alchemical binding free energy calculations for computer-aided drug design. However, parameters from transferable molecular mechanics force fields can be unreliable, and training accurate molecule-specific parameters against quantum mechanical (QM) reference data is slow. Here, we present presto (https://github.com/cole-group/presto), a Python package for fast training of bespoke SMIRKS-Native Open Force Field (SMIRNOFF) format force fields. presto generates training data using high-temperature molecular dynamics with well-tempered metadynamics. It iteratively fits all valence parameters to energies and forces from transferable machine learning potentials (MLPs). By leveraging fast MLP evaluations and training on the GPU using PyTorch (via the packages smee and descent), the fully-automated workflow completes in around 15 minutes for a 50-atom molecule on a single GPU. presto reduces torsion scan errors by a factor of three relative to Open Force Field Sage 2.3.0 on 400 molecules from TorsionNet500. It also achieves torsion scan performance comparable to direct QM fitting on a benchmark dataset of fragmented drug-like molecules, and relative conformer energies of 510 molecules from the set of Folmsbee and Hutchison are significantly improved compared to both Sage and espaloma. presto allows simultaneous fitting of congeneric series with partially shared parameters. It produces comparable relative binding free energy performance to OpenFF 1.3.1 (Parsley) on systems where Parsley already performs well, and may improve performance in cases where initial parameters are particularly poor.

Cite this paper

Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.

APA 7

Clark, F., Pope, T. J., Maier, S., Boothroyd, S., Horton, J. T., Ryczko, K., Bortolato, A., & Cole, D. J. (2026). Fast training of bespoke SMIRNOFF-format molecular mechanics force fields using machine learning potentials. ChemRxiv. https://doi.org/10.26434/chemrxiv.15004169/v2

MLA 9

Clark, Finlay, et al. “Fast training of bespoke SMIRNOFF-format molecular mechanics force fields using machine learning potentials.” ChemRxiv, 2026. https://doi.org/10.26434/chemrxiv.15004169/v2.

Chicago

Clark, Finlay, Thomas J. Pope, Sarah Maier, Simon Boothroyd, Josh T. Horton, Kevin Ryczko, Andrea Bortolato, and Daniel J. Cole. 2026. “Fast training of bespoke SMIRNOFF-format molecular mechanics force fields using machine learning potentials.” ChemRxiv. https://doi.org/10.26434/chemrxiv.15004169/v2.

Harvard

Clark, F. et al. (2026) ‘Fast training of bespoke SMIRNOFF-format molecular mechanics force fields using machine learning potentials’, ChemRxiv. doi:10.26434/chemrxiv.15004169/v2.

IEEE

F. Clark et al., “Fast training of bespoke SMIRNOFF-format molecular mechanics force fields using machine learning potentials,” ChemRxiv, 2026, doi: 10.26434/chemrxiv.15004169/v2.

BibTeX

@misc{clark2026fast,
  title     = {{Fast training of bespoke SMIRNOFF-format molecular mechanics force fields using machine learning potentials}},
  author    = {Clark, Finlay and Pope, Thomas J. and Maier, Sarah and Boothroyd, Simon and Horton, Josh T. and Ryczko, Kevin and Bortolato, Andrea and Cole, Daniel J.},
  howpublished = {ChemRxiv},
  year      = {2026},
  doi       = {10.26434/chemrxiv.15004169/v2},
  url       = {https://doi.org/10.26434/chemrxiv.15004169/v2}
}

Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)

Topic and keywords

Life Sciences › Pharmacology, Toxicology and Pharmaceutics › Drug Discovery › Computational Drug Discovery Methods

  • machine learning interatomic potentials
  • PRESTO
  • conformational sampling
  • binding free energy
  • alchemical free energy calculations
  • high-temperature molecular dynamics
  • well-tempered metadynamics
  • drug-like molecules

Work with this paper

Most-cited papers in Pharmacology, Toxicology and Pharmaceutics

All highly cited Pharmacology, Toxicology and Pharmaceutics papers

Questions about this paper

How do I cite this paper in APA style?

Clark, F., Pope, T. J., Maier, S., Boothroyd, S., Horton, J. T., Ryczko, K., Bortolato, A., & Cole, D. J. (2026). Fast training of bespoke SMIRNOFF-format molecular mechanics force fields using machine learning potentials. ChemRxiv. https://doi.org/10.26434/chemrxiv.15004169/v2

How many times has this paper been cited?

0 times, according to OpenAlex on 16 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.

Is this paper free to read?

Yes. It is open access: read the free full text at ChemRxiv.

Who wrote this paper?

Finlay Clark, Thomas J. Pope, Sarah Maier, Simon Boothroyd, Josh T. Horton, Kevin Ryczko, Andrea Bortolato, Daniel J. Cole.

What is the DOI of this paper?

10.26434/chemrxiv.15004169/v2. The permanent link is https://doi.org/10.26434/chemrxiv.15004169/v2.

Bibliographic data from OpenAlex (CC0), last checked 16 November 2026. ResearchGuru does not host this paper; the links go to the publisher or the repository that does.