Reproduce “RetroGEF: Dynamic Graph Edit Flow for Single-Step Retrosynthesis” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Xiaozhuang Song, Xuemin Chen, Xinjian Zhao, Yaoyao Xu, Tianshu Yu

RetroGEF predicts retrosynthetic precursors by growing and revising a graph through complete edits, adding an attached atom with its first bond in one update. A paired-endpoint bridge supplies training across alternative edit orders and correspondences. On USPTO-Full with predicted centers, it reaches 54.3% Top-1 and 77.1% Top-10, versus RetroDiT’s 51.3% and 75.8%; these are topology matches to recorded reactants, not experimental validation.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.