Reproduce “Retroformer: Pushing the Limits of Interpretable End-to-end Retrosynthesis Transformer” 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
Yue Wan, Benben Liao, Chang-Yu Hsieh, Shengyu Zhang
Retroformer performs template-free retrosynthesis by identifying candidate reaction centers in a product molecule and using them to guide reactant generation. Its model combines local molecular-graph attention with global SMILES attention, without using cheminformatics tools to edit molecules at inference. On USPTO-50K with reaction class unknown, Retroformer_aug+ reaches 53.2% top-1 and 82.1% top-10, below template-based LocalRetro.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.