About DecompDiff

Structure-Based Drug Design with Decomposed Diffusion Models

Cite This Tool

If you use DecompDiff in your research, please cite the original paper:

Guan, J., Zhou, X., Yang, Y., Bao, Y., Peng, J., Ma, J., Liu, Q., Wang, L. & Gu, Q. DecompDiff: Diffusion Models with Decomposed Priors for Structure-Based Drug Design. Proceedings of the 40th International Conference on Machine Learning (ICML), PMLR 202, 11827–11846 (2023).
URL: proceedings.mlr.press/v202/guan23a

BibTeX

@inproceedings{guan2023decompdiff,
  title={DecompDiff: Diffusion Models with Decomposed Priors for Structure-Based Drug Design},
  author={Guan, Jiaqi and Zhou, Xiangxin and Yang, Yuwei and Bao, Yu and Peng, Jian and Ma, Jianzhu and Liu, Qiang and Wang, Liang and Gu, Quanquan},
  booktitle={Proceedings of the 40th International Conference on Machine Learning},
  pages={11827--11846},
  year={2023},
  volume={202},
  series={Proceedings of Machine Learning Research},
  publisher={PMLR}
}

View Paper GitHub Repository

License: CC-BY-NC 4.0 — This tool is used under a non-commercial license. It may not be used for commercial purposes. Developed by ByteDance Research.

Overview

DecompDiff is a diffusion model for structure-based drug design that decomposes the ligand molecule into two parts—arms and scaffold—and applies decomposed priors for 3D molecule generation within protein binding pockets.

By decomposing the molecular generation process, DecompDiff achieves higher-quality drug-like molecules with better binding affinity and pharmacological properties compared to monolithic generation approaches. The model generates 3D molecular conformations directly within the target protein's binding pocket, producing candidates ready for downstream docking and optimization.

Key Capabilities

Decomposed Generation
Separate diffusion priors for scaffold and arms improve molecular quality
3D Pocket-Aware Design
Generates molecules directly within the target binding pocket
Bond Diffusion
Explicit bond-type generation ensures valid chemical structures
Validity Guidance
Sampling-phase guidance improves drug-likeness and synthetic accessibility

Applications

Hit Discovery
Generate novel chemical matter for undrugged or challenging targets
Lead Optimization
Explore structural modifications around known active scaffolds
Scaffold Hopping
Discover new scaffolds that maintain binding pocket complementarity
Fragment-Based Drug Design
Generate elaborated molecules from fragment hits using pocket context

Ready to Generate Drug Candidates?

Go to Dashboard New Generation Job View My Jobs