Template-Free Retrosynthesis with Graph-Prior Augmented Transformers

TL;DR

This paper introduces a template-free, Transformer-based model for retrosynthesis prediction that integrates molecular graph information and data augmentation, achieving state-of-the-art results without relying on reaction templates.

Contribution

The proposed framework removes the need for reaction templates and incorporates molecular graph data into Transformers, improving accuracy and robustness in retrosynthesis prediction.

Findings

Achieves state-of-the-art performance among template-free methods.

Substantially outperforms baseline Transformer models.

Demonstrates robustness and scalability on USPTO-50K dataset.

Abstract

Retrosynthesis reaction prediction aims to infer plausible reactant molecules for a given product and is a important problem in computer-aided organic synthesis. Despite recent progress, many existing models still fall short of the accuracy and robustness required for practical deployment. In this paper, we present a template-free, Transformer-based framework that removes the need for handcrafted reaction templates or additional chemical rule engines. Our model injects molecular graph information into the attention mechanism to jointly exploit SMILES sequences and structural cues, and further applies a paired data augmentation strategy to enhance training diversity and scale. Extensive experiments on the USPTO-50K benchmark demonstrate that our approach achieves state-of-the-art performance among template-free methods and substantially outperforms a vanilla Transformer baseline.