Skip to content
DraconiticResearch
Search Sources API
190,860 studies31 sourcesUpdated 34 min ago
PreprintCC BY-NCChemRxiv

Leveraging Machine Learning Interatomic Potentials into Fitting Fast and Accurate AMBER-Compatible Precise Force Field

  • Xuchao Zhou1
  • Xujian Wang1,2
  • Jiahao Xie1
  • Leshen Zhang3
  • Junmei Wang1
  1. 1Department of Pharmaceutical Sciences and Computational Chemical Genomics Screening Center, School of Pharmacy, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, United States
  2. 2Department of Computational and Systems Biology, School of Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, United States
  3. 3Aiiso Yufeng Li Family Department of Chemical and Nano Engineering, University of California San Diego, CA 92093, United States
Open at sourcePDFDOICite

Abstract

Transferable force fields enable routine molecular dynamics simulations across broad chemical space, but finite transferable parameter sets may not describe unusual chemical environments with uniform accuracy.Here, we present ParmFit, an automated preciseforce-field parameterization framework within MAPLE that uses machine-learning interatomic potentials to efficiently construct and refine AMBER/GAFF-compatible parameters.ParmFit extends this strategy across small molecules, chemically modified biomolecular residues, and metal coordination environments while selectively retaining compatible force field terms.In a challenging subset of FreeSolv, UMA-based precise parameters reduced the average hydration free-energy prediction error by more than 1 kcal mol -1 , while ParmFit-generated models of covalently modified proteins and metalloproteins maintained stable structural and coordination behavior in molecular dynamics simulations.By integrating efficient MLIP calculations, selective parameter refinement, and direct compatibility with established force-field workflows, ParmFit provides a practical route toward broader use of precise force fields in molecular simulation.

Related

  1. Symbolic Ensemble Learning Enables Discovery of Fast Accurate Physics-Based Interatomic Potentials
    Bilvin Varughese, Aditya Koneru, Adil Muhammad +6
    arXivPreprintCC BY-NC-SA 4.0Full textCondensed matterarXiv
  2. Universal Interatomic Potentials as Configuration-Space Generators for One-Shot and Iterative Fine-Tuning of Ab Initio-Accurate Material-Specific Models
    Jonas Hänseroth, Aaron Flötotto, Christian Dreßler
    arXivPreprintarXiv licenceFull textPhysical chemistryCondensed matterarXiv
  3. Benchmarking machine-learning interatomic potentials using classical force fields
    Johannes Sandberg, Geneviève Dusson, Julien Lam
    arXivPreprintCC BY 4.0Full textCondensed matterarXiv
  4. MLIP Detective: Active Failure Mode Discovery Beyond Benchmark Scores for Machine-Learning Interatomic Potentials
    Ryuhei Okuno, Nontawat Charoenphakdee, Kaoru Hisama +1
    arXivPreprintCC BY 4.0Full textPhysical chemistryCondensed matterarXiv
  5. Cluster-based Structural Similarity for Dataset Visualization and Data Selection for Machine Learning Interatomic Potentials
    Yuto Iwasaki, Miguel A. Caro
    arXivPreprintarXiv licenceFull textCondensed matterarXiv

Cite

BibTeXRIS
BibTeX
@misc{zhou2026leveraging,
  title = {{Leveraging Machine Learning Interatomic Potentials into Fitting Fast and Accurate AMBER-Compatible Precise Force Field}},
  author = {Xuchao Zhou and Xujian Wang and Jiahao Xie and Leshen Zhang and Junmei Wang},
  howpublished = {ChemRxiv},
  year = {2026},
  month = sep,
  doi = {10.26434/chemrxiv.15008812/v1},
  url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008812/v1},
  note = {CC BY-NC},
}
RIS
TY  - UNPB
TI  - Leveraging Machine Learning Interatomic Potentials into Fitting Fast and Accurate AMBER-Compatible Precise Force Field
AU  - Xuchao Zhou
AU  - Xujian Wang
AU  - Jiahao Xie
AU  - Leshen Zhang
AU  - Junmei Wang
PY  - 2026
DA  - 2026/09/14
PB  - ChemRxiv
DO  - 10.26434/chemrxiv.15008812/v1
UR  - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15008812/v1
AB  - Transferable force fields enable routine molecular dynamics simulations across broad chemical space, but finite transferable parameter sets may not describe unusual chemical environments with uniform accuracy.Here, we present ParmFit, an automated preciseforce-field parameterization framework within MAPLE that uses machine-learning interatomic potentials to efficiently construct and refine AMBER/GAFF-compatible parameters.ParmFit extends this strategy across small molecules, chemically modified biomolecular residues, and metal coordination environments while selectively retaining compatible force field terms.In a challenging subset of FreeSolv, UMA-based precise parameters reduced the average hydration free-energy prediction error by more than 1 kcal mol -1 , while ParmFit-generated models of covalently modified proteins and metalloproteins maintained stable structural and coordination behavior in molecular dynamics simulations.By integrating efficient MLIP calculations, selective parameter refinement, and direct compatibility with established force-field workflows, ParmFit provides a practical route toward broader use of precise force fields in molecular simulation.
ER  -