Order Matters in Retrosynthesis: Structure-aware Generation via Reaction-Center-Guided Discrete Flow Matching

TL;DR

This paper introduces a structure-aware, template-free retrosynthesis framework that leverages atom ordering and reaction center positioning to improve prediction accuracy and efficiency, outperforming existing models with less data and fewer steps.

Contribution

It proposes a novel structure-aware approach using reaction-center-guided positional encoding and a graph transformer backbone for more accurate and efficient retrosynthesis prediction.

Findings

Achieves state-of-the-art top-1 accuracy on USPTO datasets.

Generates predictions in 20-50 steps, significantly fewer than prior diffusion methods.

Structural priors outperform brute-force scaling, enabling smaller models to match larger ones.

Abstract

Template-free retrosynthesis methods treat the task as black-box sequence generation, limiting learning efficiency, while semi-template approaches rely on rigid reaction libraries that constrain generalization. We address this gap with a key insight: atom ordering in neural representations matters. Building on this insight, we propose a structure-aware template-free framework that encodes the two-stage nature of chemical reactions as a positional inductive bias. By placing reaction center atoms at the sequence head, our method transforms implicit chemical knowledge into explicit positional patterns that the model can readily capture. The proposed RetroDiT backbone, a graph transformer with rotary position embeddings, exploits this ordering to prioritize chemically critical regions. Combined with discrete flow matching, our approach decouples training from sampling and enables generation in 20--50 steps versus 500 for prior diffusion methods. Our method achieves state-of-the-art performance on both USPTO-50k (61.2% top-1) and the large-scale USPTO-Full (51.3% top-1) with predicted reaction centers. With oracle centers, performance reaches 71.1% and 63.4% respectively, surpassing foundation models trained on 10 billion reactions while using orders of magnitude less data. Ablation studies further reveal that structural priors outperform brute-force scaling: a 280K-parameter model with proper ordering matches a 65M-parameter model without it.