EUVIMEDEUVIMED entdecken
Open Access

Steering semi-flexible molecular diffusion model for structure-based drug design with reinforcement learning

Xudong Zhang iD, Sanqing Qu, Fan Lu iD, Jianmin Wang iD, Zhixin Tian iD, Shangding Gu, Yanping Zhang iD, Alois Knoll iD, Shaorong Gao iD, Guang Chen iD, Changjun Jiang

Science Advances · 17.04.2026 · 12 (16)

DOI: 10.1126/sciadv.ady9955 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Science Advances, erschienen am 17.04.2026. Verfasst von Xudong Zhang und 10 weiteren Autor:innen. Beteiligte Einrichtung: Shanghai Key Laboratory of Maternal Fetal Medicine, Clinical and Translational Research Center of Shanghai First Maternity and Infant Hospital, Schoo…. Hauptthemen laut MEDLINE: Drug Design, Models, Molecular, Proteins. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 62 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar. Gefördert von: National Natural Science Foundation of China, China Postdoctoral Science Foundation, National Key Research and Development Program of China.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Current structure-based molecular generation faces a fundamental dilemma: While static ligand modeling dominates computational approaches, real-world molecular interactions are inherently dynamic. Inspired by the conformational changes ligands undergo during semi-flexible docking, we propose a reinforcement learning (RL)–steered diffusion framework for semi-flexible molecular generation in protein pockets. By defining the denoising process as a Markov decision process, RL dynamically adjusts molecular structures through iterative exploration. Simultaneously, we incorporate multiple molecular properties as conditions to constrain the denoising policy to drug-like regions and perform self-supervised rigid training on both target-free and target-specific molecules. In addition, we propose a fast sampling strategy that accelerates sampling by 20 times, thereby improving the efficiency of training and sampling. Experiments demonstrate that our method outperforms state-of-the-art methods with a Vina score of −7.23 kcal/mol and an 11.53% success rate. Targeting unseen real-world proteins, the generated molecules preserve canonical interaction patterns while discovering previously unknown binding chemotypes.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Drug DesignModels, MolecularProteinsReinforcement Machine LearningLigandsMolecular Docking SimulationMarkov ChainsAlgorithms

Substanzen

LigandsProteins

Autor:innen und Institutionen

  • Shanghai Key Laboratory of Maternal Fetal Medicine, Clinical and Translational Research Center of Shanghai First Maternity and Infant Hospital, School of Computer Science and Technology, Tongji University, Shanghai 2000…
  • Shanghai Innovation Institute, Shanghai, China.
  • MOE Key Laboratory of Embedded System and Service Computing, School of Computer Science and Technology, Tongji University, Shanghai, China.
  • The Interdisciplinary Graduate Program in Integrative Biotechnology, Yonsei University, Incheon, Republic of Korea.
  • School of Chemical Science & Engineering and Shanghai Key Laboratory of Chemical Assessment and Sustainability, Tongji University, Shanghai, China.
  • Department of Electrical Engineering and Computer Sciences, University of California, Berkeley, CA, USA.

ORCID-Profile: Xudong Zhang, Fan Lu, Jianmin Wang, Zhixin Tian, Yanping Zhang, Alois Knoll, Shaorong Gao, Guang Chen

Förderung

  • National Natural Science Foundation of China
  • China Postdoctoral Science Foundation
  • National Key Research and Development Program of China
  • European Union’s Horizon 2020 Framework Programme
  • Xiaomi Young Talents Program

Zitieren

APA

Zhang, X., Qu, S., Lu, F., Wang, J., Tian, Z., Gu, S., Zhang, Y., Knoll, A., Gao, S., Chen, G., & Jiang, C. (2026). Steering semi-flexible molecular diffusion model for structure-based drug design with reinforcement learning. Science Advances, 12(16). https://doi.org/10.1126/sciadv.ady9955

Vancouver

Zhang X, Qu S, Lu F, Wang J, Tian Z, Gu S, et al. Steering semi-flexible molecular diffusion model for structure-based drug design with reinforcement learning. Science Advances. 2026;12(16). doi:10.1126/sciadv.ady9955

BibTeX
@article{Zhang2026,
  author = {Zhang, Xudong and Qu, Sanqing and Lu, Fan and Wang, Jianmin and Tian, Zhixin and Gu, Shangding and Zhang, Yanping and Knoll, Alois and Gao, Shaorong and Chen, Guang and Jiang, Changjun},
  title = {Steering semi-flexible molecular diffusion model for structure-based drug design with reinforcement learning},
  journal = {Science Advances},
  year = {2026},
  volume = {12},
  number = {16},
  doi = {10.1126/sciadv.ady9955}
}
Kostenlos starten →