EUVIMEDEUVIMED entdecken
Crossref

MolGuidance: Advanced Guidance Strategies for Conditional Molecular Generation with Flow Matching

Jirui Jin, Cheng Zeng iD, Pawan Prakash, Ellad B Tadmor iD, Adrian Roitberg iD, Richard G Hennig iD, Stefano Martiniani, Mingjie Liu iD

Journal of Chemical Information and Modeling · 10.08.2026 · 66 (15) · 8860-8874

DOI: 10.1021/acs.jcim.6c00964 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Chemical Information and Modeling, online erschienen am 21.07.2026. Verfasst von Jirui Jin und 7 weiteren Autor:innen. Beteiligte Einrichtung: Department of Chemistry, University of Florida, Gainesville, Florida32611, United States. Hauptthemen laut MEDLINE: Models, Chemical. Die Arbeit nennt 51 Literaturangaben. Gefördert von: Division of Advanced Cyberinfrastructure.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Key objectives in conditional molecular generation include ensuring chemical validity, aligning generated molecules with target properties, promoting structural diversity, and enabling efficient sampling for discovery. Recent advances in computer vision introduced a range of new guidance strategies for generative models, many of which can be adapted to support these goals. In this work, we integrate state-of-the-art guidance methods─including classifier-free guidance, autoguidance, and model guidance─in a leading molecule generation framework built on an SE(3)-equivariant flow matching process. We propose a hybrid guidance strategy that separately guides continuous and discrete molecular modalities─operating on velocity fields and predicted denoising probabilities, respectively─while jointly optimizing their guidance scales via Bayesian optimization. Our implementation, benchmarked on the QM9 and QMe14S data sets, achieves new state-of-the-art performance in property alignment for de novo molecular generation. The generated molecules also exhibit high structural validity. Furthermore, we systematically compare the strengths and limitations of various guidance methods, offering insights into their broader applicability.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Models, ChemicalBayes TheoremGenerative Artificial Intelligence

Autor:innen und Institutionen

  • Department of Chemistry, University of Florida, Gainesville, Florida32611, United States.
  • Quantum Theory Project, University of Florida, Gainesville, Florida32611, United States.
  • Department of Physics, University of Florida, Gainesville, Florida32611, United States.
  • Department of Aerospace Engineering and Mechanics, University of Minnesota, Minneapolis, Minnesota55455, United States.
  • Department of Materials Science & Engineering, University of Florida, Gainesville, Florida32611, United States.
  • Center for Soft Matter Research, Department of Physics, New York University, New York, New York10003, United States.

ORCID-Profile: Cheng Zeng, Ellad B Tadmor, Adrian Roitberg, Richard G Hennig, Mingjie Liu

Förderung

  • Division of Advanced Cyberinfrastructure

Zitieren

APA

Jin, J., Zeng, C., Prakash, P., Tadmor, E. B., Roitberg, A., Hennig, R. G., Martiniani, S., & Liu, M. (2026). MolGuidance: Advanced Guidance Strategies for Conditional Molecular Generation with Flow Matching. Journal of Chemical Information and Modeling, 66(15), 8860-8874. https://doi.org/10.1021/acs.jcim.6c00964

Vancouver

Jin J, Zeng C, Prakash P, Tadmor EB, Roitberg A, Hennig RG, et al. MolGuidance: Advanced Guidance Strategies for Conditional Molecular Generation with Flow Matching. Journal of Chemical Information and Modeling. 2026;66(15):8860-8874. doi:10.1021/acs.jcim.6c00964

BibTeX
@article{Jin2026,
  author = {Jin, Jirui and Zeng, Cheng and Prakash, Pawan and Tadmor, Ellad B and Roitberg, Adrian and Hennig, Richard G and Martiniani, Stefano and Liu, Mingjie},
  title = {MolGuidance: Advanced Guidance Strategies for Conditional Molecular Generation with Flow Matching},
  journal = {Journal of Chemical Information and Modeling},
  year = {2026},
  volume = {66},
  number = {15},
  pages = {8860--8874},
  doi = {10.1021/acs.jcim.6c00964}
}
Kostenlos starten →