Retro-Expert: Collaborative Reasoning for Interpretable Retrosynthesis

2025-08-14

AI Summary

Inverse synthesis prediction suffers from weak logical reasoning capabilities and poor interpretability in existing models. To address this, we propose an interpretable collaborative reasoning framework that synergistically integrates domain-specific chemical models—performing shallow, rule-based inference—with large language models—enabling deep logical deduction—while incorporating reinforcement learning to optimize the multi-stage reasoning policy. The resulting dynamic, hierarchical architecture generates natural-language reasoning traces aligned with expert chemical intuition, ensuring transparent, knowledge-grounded decision-making. Experimental results demonstrate that our method outperforms state-of-the-art large language models and specialized chemical models on key metrics—including Top-1 accuracy—achieving significant improvements in both predictive performance and interpretability on benchmark datasets such as USPTO. This advances the alignment between AI-driven predictions and practical synthetic planning, effectively bridging the gap between computational prediction and laboratory implementation.

Technology Category

• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Retrosynthesis prediction aims to infer the reactant molecule based on a given product molecule, which is a fundamental task in chemical synthesis. However, existing models rely on static pattern-matching paradigm, which limits their ability to perform effective logic decision-making, leading to black-box decision-making. Building on this, we propose Retro-Expert, an interpretable retrosynthesis framework that performs collaborative reasoning by combining the complementary reasoning strengths of Large Language Models and specialized models via reinforcement learning. It outputs natural language explanations grounded in chemical logic through three components: (1) specialized models perform shallow reasoning to construct high-quality chemical decision space, (2) LLM-driven critical reasoning to generate predictions and corresponding interpretable reasoning path, and (3) reinforcement learning optimizing interpretable decision policy. Experiments show that Retro-Expert not only surpasses both LLM-based and specialized models across different metrics but also provides expert-aligned explanations that bridge the gap between AI predictions and actionable chemical insights.

Problem

• Overcoming static pattern-matching limitations in retrosynthesis prediction
• Enhancing interpretability of AI-driven chemical synthesis decisions
• Integrating LLMs and specialized models for collaborative reasoning

Innovation

• Combines LLMs and specialized models via reinforcement learning
• Generates interpretable reasoning paths with natural language explanations
• Optimizes decision policy through reinforcement learning for accuracy