EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

SynLlama: Generating Synthesizable Molecules and Their Analogs with Large Language Models

Kunyang Sun iD, Dorian Bagni, Joseph M Cavanagh, Yingze Wang, Jacob M Sawyer, Bo Zhou iD, Andrew Gritsevskiy, Oufan Zhang, Teresa Head-Gordon iD

ACS Central Science · 26.11.2025 · 11 (11) · 2108-2120

DOI: 10.1021/acscentsci.5c01285 ↗

Kurz gefasst

Fachartikel in der Zeitschrift ACS Central Science, online erschienen am 17.09.2025. Verfasst von Kunyang Sun und 8 weiteren Autor:innen. Beteiligte Einrichtung: †Kenneth S. Pitzer Theory Center and Department of Chemistry, ‡Department of Bioengineering, and §Department of Chemical and Biomolecular Engineering…. Die Arbeit wurde bisher 22-mal zitiert (Crossref) und nennt 80 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: NIAID NIH HHS, National Institute of Allergy and Infectious Diseases.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Generative machine learning models for exploring chemical space have shown immense promise, but many molecules that they generate are too difficult to synthesize, making them impractical for further investigation or development. In this work, we present a novel approach by fine-tuning Meta's Llama3 Large Language Models (LLMs) to create SynLlama, which generates full synthetic pathways made of commonly accessible building blocks and robust organic reaction templates. SynLlama explores a large synthesizable space using significantly less data and offers strong performance in both forward and bottom-up synthesis planning compared to other state-of-the-art methods. We find that SynLlama, even without training on external building blocks, can effectively generalize to unseen yet purchasable building blocks, meaning that its reconstruction capabilities extend to a broader synthesizable chemical space than those of the training data. We also demonstrate the use of SynLlama in a pharmaceutical context for synthesis planning of analog molecules and hit expansion leads for proposed inhibitors of target proteins, offering medicinal chemists a valuable tool for discovery.

Autor:innen und Institutionen

  • †Kenneth S. Pitzer Theory Center and Department of Chemistry, ‡Department of Bioengineering, and §Department of Chemical and Biomolecular Engineering, University of California, Berkeley, California 94720, United States.
  • Department of Chemistry, University of Minnesota, 207 Pleasant Street SE, Minneapolis, Minnesota 55455, United States.
  • Department of Pharmaceutical Sciences, University of Illinois Chicago, 833 S Wood St, Chicago, Illinois 60612, United States.
  • Contramont Research, San Francisco, California 94158, United States.

ORCID-Profile: Kunyang Sun, Bo Zhou, Teresa Head-Gordon

Förderung

  • NIAID NIH HHS
  • National Institute of Allergy and Infectious Diseases

Zitieren

APA

Sun, K., Bagni, D., Cavanagh, J. M., Wang, Y., Sawyer, J. M., Zhou, B., Gritsevskiy, A., Zhang, O., & Head-Gordon, T. (2025). SynLlama: Generating Synthesizable Molecules and Their Analogs with Large Language Models. ACS Central Science, 11(11), 2108-2120. https://doi.org/10.1021/acscentsci.5c01285

Vancouver

Sun K, Bagni D, Cavanagh JM, Wang Y, Sawyer JM, Zhou B, et al. SynLlama: Generating Synthesizable Molecules and Their Analogs with Large Language Models. ACS Central Science. 2025;11(11):2108-2120. doi:10.1021/acscentsci.5c01285

BibTeX
@article{Sun2025,
  author = {Sun, Kunyang and Bagni, Dorian and Cavanagh, Joseph M and Wang, Yingze and Sawyer, Jacob M and Zhou, Bo and Gritsevskiy, Andrew and Zhang, Oufan and Head-Gordon, Teresa},
  title = {SynLlama: Generating Synthesizable Molecules and Their Analogs with Large Language Models},
  journal = {ACS Central Science},
  year = {2025},
  volume = {11},
  number = {11},
  pages = {2108--2120},
  doi = {10.1021/acscentsci.5c01285}
}
Kostenlos starten →