2. Academic Validation
  3. Exploring the macrocyclic chemical space for heuristic drug design with deep learning models

Exploring the macrocyclic chemical space for heuristic drug design with deep learning models

  • Commun Chem. 2025 Oct 7;8(1):299. doi: 10.1038/s42004-025-01686-w.
Feng Hu # 1,  Xiaotong Jia # 1,  Wenjie Liao # 1,  Ziqi Chen 1,  Hongjie Bi 1,  Huan Ge 1,  Dandan Liu 1,  Rongrong Zhang 1,  Yuting Hu 1,  Wenyi Mei 1,  Zhenjiang Zhao 1,  Kai Zhang 2,  Lili Zhu 3,  Yanyan Diao 4,  Honglin Li 5 6
Affiliations

Affiliations

  • 1 Shanghai Key Laboratory of New Drug Design, School of Pharmacy, East China University of Science and Technology, Shanghai, China.
  • 2 Innovation Center for AI and Drug Discovery, School of Pharmacy, East China Normal University, Shanghai, China. kzhang@cs.ecnu.edu.cn.
  • 3 Shanghai Key Laboratory of New Drug Design, School of Pharmacy, East China University of Science and Technology, Shanghai, China. zhulfl@ecust.edu.cn.
  • 4 Innovation Center for AI and Drug Discovery, School of Pharmacy, East China Normal University, Shanghai, China. diaoyan1025@126.com.
  • 5 Shanghai Key Laboratory of New Drug Design, School of Pharmacy, East China University of Science and Technology, Shanghai, China. hlli@ecust.edu.cn.
  • 6 Innovation Center for AI and Drug Discovery, School of Pharmacy, East China Normal University, Shanghai, China. hlli@ecust.edu.cn.
  • # Contributed equally.
PMID: 41057501 DOI: 10.1038/s42004-025-01686-w
Abstract

Macrocyclic compounds hold great promise as therapeutic agents. However, their structural optimization remains constrained by the limited availability of bioactive candidates, which in turn hampers the systematic exploration of structure-activity relationships. Here we introduce CycleGPT, a generative chemical language model designed specifically to address these challenges. CycleGPT is characterized by a progressive transfer learning paradigm that incrementally transfers knowledge from pre-trained chemical language models to specialized macrocycle generation, thereby overcoming the data shortage issue. Meanwhile, it adopts an innovative probabilistic sampling strategy that effectively improves the structural novelty of generated macrocycles while ensuring domain-specific adaptability. In a prospective drug design based on CycleGPT and a JAK2 activity prediction model, we successfully developed a new JAK2 drug candidate with a good selectivity profile (inhibiting 17 wild-type kinases) and promising potential for treating polycythemia in vivo, demonstrating the practicality of deep learning methods in macrocyclic drug design.

Figures
Products
嗨！很高兴为您提供帮助！

尊敬的 MCE 客户您好，
请您选择所在区域，我们将转接对应客服为您服务！

热门区域

A

B

C

F

G

H

J

L

N

Q

S

T

X

Y

Z

热 A B C F G H J L N Q S T X Y Z