RetroDiff: Retrosynthesis as Multi-stage Distribution Interpolation

Yiming Wang, Yuxuan Song, Yiqun Wang, Minkai Xu, Rui Wang, Hao Zhou, Wei-Ying Ma
Proceedings of The 28th International Conference on Artificial Intelligence and Statistics, PMLR 258:1918-1926, 2025.

Abstract

Retrosynthesis poses a key challenge in biopharmaceuticals, aiding chemists in finding appropriate reactant molecules for given product molecules. With reactants and products represented as 2D graphs, retrosynthesis constitutes a conditional graph-to-graph (G2G) generative task. Inspired by advancements in discrete diffusion models for graph generation, we aim to design a diffusion-based method to address this problem. However, integrating a diffusion-based G2G framework while retaining essential chemical reaction template information presents a notable challenge. Our key innovation involves a multi-stage diffusion process. We decompose the retrosynthesis procedure to first sample external groups from the dummy distribution given products, then generate external bonds to connect products and generated groups. Interestingly, this generation process mirrors the reverse of the widely adapted semi-template retrosynthesis workflow, i.e., from reaction center identification to synthon completion. Based on these designs, we introduce Retrosynthesis Diffusion (RetroDiff), a novel diffusion-based method for the retrosynthesis task. Experimental results demonstrate that RetroDiff surpasses all semi-template methods in accuracy, and outperforms template-based and template-free methods in large-scale scenarios and molecular validity, respectively.

Cite this Paper

BibTeX
@InProceedings{pmlr-v258-wang25e, title = {RetroDiff: Retrosynthesis as Multi-stage Distribution Interpolation}, author = {Wang, Yiming and Song, Yuxuan and Wang, Yiqun and Xu, Minkai and Wang, Rui and Zhou, Hao and Ma, Wei-Ying}, booktitle = {Proceedings of The 28th International Conference on Artificial Intelligence and Statistics}, pages = {1918--1926}, year = {2025}, editor = {Li, Yingzhen and Mandt, Stephan and Agrawal, Shipra and Khan, Emtiyaz}, volume = {258}, series = {Proceedings of Machine Learning Research}, month = {03--05 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v258/main/assets/wang25e/wang25e.pdf}, url = {https://proceedings.mlr.press/v258/wang25e.html}, abstract = {Retrosynthesis poses a key challenge in biopharmaceuticals, aiding chemists in finding appropriate reactant molecules for given product molecules. With reactants and products represented as 2D graphs, retrosynthesis constitutes a conditional graph-to-graph (G2G) generative task. Inspired by advancements in discrete diffusion models for graph generation, we aim to design a diffusion-based method to address this problem. However, integrating a diffusion-based G2G framework while retaining essential chemical reaction template information presents a notable challenge. Our key innovation involves a multi-stage diffusion process. We decompose the retrosynthesis procedure to first sample external groups from the dummy distribution given products, then generate external bonds to connect products and generated groups. Interestingly, this generation process mirrors the reverse of the widely adapted semi-template retrosynthesis workflow, i.e., from reaction center identification to synthon completion. Based on these designs, we introduce Retrosynthesis Diffusion (RetroDiff), a novel diffusion-based method for the retrosynthesis task. Experimental results demonstrate that RetroDiff surpasses all semi-template methods in accuracy, and outperforms template-based and template-free methods in large-scale scenarios and molecular validity, respectively.} }
Endnote
%0 Conference Paper %T RetroDiff: Retrosynthesis as Multi-stage Distribution Interpolation %A Yiming Wang %A Yuxuan Song %A Yiqun Wang %A Minkai Xu %A Rui Wang %A Hao Zhou %A Wei-Ying Ma %B Proceedings of The 28th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2025 %E Yingzhen Li %E Stephan Mandt %E Shipra Agrawal %E Emtiyaz Khan %F pmlr-v258-wang25e %I PMLR %P 1918--1926 %U https://proceedings.mlr.press/v258/wang25e.html %V 258 %X Retrosynthesis poses a key challenge in biopharmaceuticals, aiding chemists in finding appropriate reactant molecules for given product molecules. With reactants and products represented as 2D graphs, retrosynthesis constitutes a conditional graph-to-graph (G2G) generative task. Inspired by advancements in discrete diffusion models for graph generation, we aim to design a diffusion-based method to address this problem. However, integrating a diffusion-based G2G framework while retaining essential chemical reaction template information presents a notable challenge. Our key innovation involves a multi-stage diffusion process. We decompose the retrosynthesis procedure to first sample external groups from the dummy distribution given products, then generate external bonds to connect products and generated groups. Interestingly, this generation process mirrors the reverse of the widely adapted semi-template retrosynthesis workflow, i.e., from reaction center identification to synthon completion. Based on these designs, we introduce Retrosynthesis Diffusion (RetroDiff), a novel diffusion-based method for the retrosynthesis task. Experimental results demonstrate that RetroDiff surpasses all semi-template methods in accuracy, and outperforms template-based and template-free methods in large-scale scenarios and molecular validity, respectively.
APA
Wang, Y., Song, Y., Wang, Y., Xu, M., Wang, R., Zhou, H. & Ma, W.. (2025). RetroDiff: Retrosynthesis as Multi-stage Distribution Interpolation. Proceedings of The 28th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 258:1918-1926 Available from https://proceedings.mlr.press/v258/wang25e.html.

Related Material