Compositional Flows for 3D Molecule and Synthesis Pathway Co-design
Tony Shen, Seonghwan Seo, Ross Irwin, Kieran Didi, Simon Olsson, Woo Youn Kim, Martin Ester · 2025
Read Compositional Flows for 3D Molecule and Synthesis Pathway Co-design by Tony Shen, Seonghwan Seo, Ross Irwin, Kieran Didi, Simon Olsson, Woo Youn Kim, Martin Ester (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-08051.txt.
Description
Tony Shen and co-authors teach a generative system to design a drug and its recipe at the same time. A molecule that binds well is useless if chemists cannot make it. Their framework, CGFlow, grows the synthesis pathway step by step. At the same time it shapes the molecule's three-dimensional pose in its target pocket.
The method extends flow matching, a generative technique, to objects built in pieces with continuous features, and it borrows theory from generative flow networks to steer sampling toward high-reward designs. On a standard benchmark of fifteen targets, it reached the best binding scores across the board, and it sampled nearly six times more efficiently than a two-dimensional synthesis-based rival. It also led on docking score and on the share of designs a synthesis planner could actually build.
Catalog record
| CREATOR | Tony Shen, Seonghwan Seo, Ross Irwin, Kieran Didi, Simon Olsson, Woo Youn Kim, Martin Ester |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 95 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-08051