Synthelite: Chemist-aligned and feasibility-aware synthesis planning with LLMs

2025-12-18

AI Summary

Existing computer-aided synthesis planning (CASP) systems lack human-in-the-loop mechanisms, hindering integration of expert chemists’ domain knowledge. Method: We propose the first large language model (LLM)-native retrosynthetic framework explicitly designed for synthetic chemists, enabling natural-language interaction and real-time expert intervention to support end-to-end, constraint-adaptive route design. Our approach unifies expert intent interpretation, chemical feasibility modeling, and zero-shot/few-shot retrosynthetic reasoning within a single LLM architecture, augmented by constraint-aware path search and posterior feasibility validation. Contribution/Results: On dual-constrained tasks—incorporating both strategic and starting-material constraints—the framework achieves a 95% success rate, markedly improving route executability and chemical plausibility. This work constitutes the first empirical demonstration that LLMs can serve as central coordinators in synthesis planning, bridging high-level strategic reasoning with low-level chemical validity.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Computer-aided synthesis planning (CASP) has long been envisioned as a complementary tool for synthetic chemists. However, existing frameworks often lack mechanisms to allow interaction with human experts, limiting their ability to integrate chemists' insights. In this work, we introduce Synthelite, a synthesis planning framework that uses large language models (LLMs) to directly propose retrosynthetic transformations. Synthelite can generate end-to-end synthesis routes by harnessing the intrinsic chemical knowledge and reasoning capabilities of LLMs, while allowing expert intervention through natural language prompts. Our experiments demonstrate that Synthelite can flexibly adapt its planning trajectory to diverse user-specified constraints, achieving up to 95% success rates in both strategy-constrained and starting-material-constrained synthesis tasks. Additionally, Synthelite exhibits the ability to account for chemical feasibility during route design. We envision Synthelite to be both a useful tool and a step toward a paradigm where LLMs are the central orchestrators of synthesis planning.

Problem

• Develops a synthesis planning framework using LLMs for retrosynthetic transformations
• Enables expert intervention via natural language prompts in route generation
• Adapts to user constraints and chemical feasibility in synthesis tasks

Innovation

• Uses LLMs to propose retrosynthetic transformations directly
• Allows expert intervention via natural language prompts
• Adapts planning to user constraints and chemical feasibility