Reasoning-Driven Retrosynthesis Prediction with Large Language Models via Reinforcement Learning

TL;DR

This paper introduces RetroDFM-R, a reasoning-based large language model for chemical retrosynthesis that uses reinforcement learning to improve accuracy, explainability, and practical utility in predicting synthetic routes.

Contribution

The paper presents RetroDFM-R, a novel LLM that leverages chemically guided reinforcement learning to enhance retrosynthesis prediction accuracy and interpretability.

Findings

Achieves 65.0% top-1 accuracy on USPTO-50K benchmark.

Outperforms state-of-the-art methods in retrosynthesis prediction.

Provides human-interpretable reasoning for predictions.

Abstract

Retrosynthesis planning, essential in organic synthesis and drug discovery, has greatly benefited from recent AI-driven advancements. Nevertheless, existing methods frequently face limitations in both applicability and explainability. Traditional graph-based and sequence-to-sequence models often lack generalized chemical knowledge, leading to predictions that are neither consistently accurate nor easily explainable. To address these challenges, we introduce RetroDFM-R, a reasoning-based large language model (LLM) designed specifically for chemical retrosynthesis. Leveraging large-scale reinforcement learning guided by chemically verifiable rewards, RetroDFM-R significantly enhances prediction accuracy and explainability. Comprehensive evaluations demonstrate that RetroDFM-R significantly outperforms state-of-the-art methods, achieving a top-1 accuracy of 65.0% on the USPTO-50K benchmark. Double-blind human assessments further validate the chemical plausibility and practical utility of RetroDFM-R's predictions. RetroDFM-R also accurately predicts multistep retrosynthetic routes reported in the literature for both real-world drug molecules and perovskite materials. Crucially, the model's explicit reasoning process provides human-interpretable insights, thereby enhancing trust and practical value in real-world retrosynthesis applications.