Skip to content
DraconiticResearch
Search Sources API
186,123 studies31 sourcesUpdated 47 min ago
Peer-reviewedCC BY 4.0Journal of Chemical Information and Modeling

Toward Autonomous Extraction and Building of Machine-Readable Molecular Models from Publications Using Large Language Models

  • Florian Fleckenstein1
  • Volodymyr Rodin2
  • Isabell Dieudonné1
  • Fadi Al Machot3
  • Silvia Chiacchiera4
  • Martin Horsch3
  • Amila Akagic2
  • Simon Stephan1
4 affiliations
  1. 1Laboratory of Heat- and Mass Transfer, Otto von Guericke University Magdeburg, 39106 Magdeburg, Germany
  2. 2Department of Computer Science and Informatics, University of Sarajevo, 71000 Sarajevo, Bosnia and Herzegovina
  3. 3Department of Data Science, Norwegian University of Life Sciences, 1432 Ås, Norway
  4. 4UKRI Science and Technology Facilities Council, Daresbury Laboratory, WA4 4AD Warrington, U.K.
Read full textPDFDOICite

Abstract

Force field models are one of the central pillars of molecular simulations. A large number of molecular force field models has been developed in the past decadesmostly published in scientific papers. Building machine-readable force field input files for simulation engines is a tedious and error-prone task. We developed a method for extracting and building force field files from publications using large language models (LLMs). We have tested the new method by extracting 114 force field models from 21 scientific publications. The studied force fields comprise 6–74 parameters that were automatically extracted, identified, annotated, and allocated. We have compared the performance of different LLMs, namely Gemini 2.5 Pro, Claude 4 Sonnet, Claude 3.7 Sonnet, Gemini 2.5 Flash, and GPT-4o. Overall, they yield similar performanceyet there are important differences in individual cases. The overall best performance was obtained by the Gemini 2.5 Pro LLM. The force field parameters were extracted and identified with an accuracy of 94.0% using the Gemini 2.5 Pro LLM. The new LLM-assisted extraction method drastically reduces the time required for building force field files. Nevertheless, human-in-the-loop verification is still indispensable.

Related

  1. Automated Extraction and FAIRification of Chemical Structures from Scientific Publications
    Felix Bänsch, Kohulan Rajan, Markus Nietfeld
    ChemRxivPreprintCC BYFull textGeneral chemistryChemRxiv
  2. PULSES: multi-agent AI for autonomous excited-state model building and dynamics
    Fangning Ren, Lechen Dong, Xu Chen +3
    arXivPreprintCC BY-NC-ND 4.0Full textPhysical chemistryarXiv
  3. C-MAGE: Automated Extraction and Organization Engine for Molecular Representations in Chemical Documents
    Alexander Taylor, Jeehyun Hwang, Mihir Surve +2
    ChemRxivPreprintLicence not statedGeneral chemistryChemRxiv
  4. Green Extraction and Molecular Docking-Guided Discovery of Asiaticoside as a Multi-Target Modulator of the SIRT1/HMGB1/p53-p21 Axis in H2O2-Induced Senescent Endothelial Cells
    Wenkai Li, Jianhua Huang, Fei Zeng +4
    Molecules (Basel, Switzerland)Peer-reviewedCC BYFull textGeneral chemistryEurope PMC
  5. Engineering a water-compatible upconversion nanoink for NIR-readable covert printing and QR authentication
    Najla Alsofyani, Anfal Alfahd, Faisal Alghannam +5
    Scientific ReportsPeer-reviewedCC BY-NC-ND 4.0Full textMaterials & polymersCrossref

Cite

BibTeXRIS
BibTeX
@article{fleckenstein2026toward,
  title = {{Toward Autonomous Extraction and Building of Machine-Readable Molecular Models from Publications Using Large Language Models}},
  author = {Florian Fleckenstein and Volodymyr Rodin and Isabell Dieudonné and Fadi Al Machot and Silvia Chiacchiera and Martin Horsch and Amila Akagic and Simon Stephan},
  journal = {Journal of Chemical Information and Modeling},
  year = {2026},
  month = oct,
  doi = {10.1021/acs.jcim.6c02128},
  url = {https://doi.org/10.1021/acs.jcim.6c02128},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - Toward Autonomous Extraction and Building of Machine-Readable Molecular Models from Publications Using Large Language Models
AU  - Florian Fleckenstein
AU  - Volodymyr Rodin
AU  - Isabell Dieudonné
AU  - Fadi Al Machot
AU  - Silvia Chiacchiera
AU  - Martin Horsch
AU  - Amila Akagic
AU  - Simon Stephan
PY  - 2026
DA  - 2026/10/06
JO  - Journal of Chemical Information and Modeling
DO  - 10.1021/acs.jcim.6c02128
UR  - https://doi.org/10.1021/acs.jcim.6c02128
AB  - Force field models are one of the central pillars of molecular simulations. A large number of molecular force field models has been developed in the past decadesmostly published in scientific papers. Building machine-readable force field input files for simulation engines is a tedious and error-prone task. We developed a method for extracting and building force field files from publications using large language models (LLMs). We have tested the new method by extracting 114 force field models from 21 scientific publications. The studied force fields comprise 6–74 parameters that were automatically extracted, identified, annotated, and allocated. We have compared the performance of different LLMs, namely Gemini 2.5 Pro, Claude 4 Sonnet, Claude 3.7 Sonnet, Gemini 2.5 Flash, and GPT-4o. Overall, they yield similar performanceyet there are important differences in individual cases. The overall best performance was obtained by the Gemini 2.5 Pro LLM. The force field parameters were extracted and identified with an accuracy of 94.0% using the Gemini 2.5 Pro LLM. The new LLM-assisted extraction method drastically reduces the time required for building force field files. Nevertheless, human-in-the-loop verification is still indispensable.
SN  - 1549-9596
SN  - 1549-960X
ER  -