MolE-RAG: Molecular Structure-Enhanced Retrieval-Augmented Generation for Chemistry

2026-06-04

AI Summary

Large language models (LLMs) struggle to effectively interpret chemical structural information encoded in non-natural language formats such as SMILES, limiting their performance in molecular property prediction. To address this challenge, this work proposes MolE-RAG, a training-free retrieval-augmented generation framework that introduces, for the first time, a molecule-centric, multi-source heterogeneous knowledge retrieval mechanism. During inference, MolE-RAG dynamically integrates three types of contextual information: chemical literature, molecular metadata (e.g., synonyms, functional groups, physicochemical descriptors), and structurally similar molecules. Without requiring fine-tuning, this approach flexibly combines structural, textual, and physicochemical knowledge, significantly enhancing the performance of general-purpose LLMs across nine molecular property prediction tasks—achieving up to a 28-percentage-point improvement in ROC-AUC for classification tasks and up to a 67% reduction in RMSE for regression tasks.

Abstract

Large language models (LLMs) have shown promise for molecular property prediction, but their ability to reason over chemical structures remains limited, as molecular representations such as SMILES differ substantially from the natural language on which LLMs are primarily trained. To bridge this semantic and chemical knowledge gap, we propose MolE-RAG, a training-free, molecule-centric retrieval-augmented generation framework for LLM-based molecular property prediction. MolE-RAG augments each prediction with three complementary sources of inference-time context: retrieved chemistry literature, molecule-specific information including compound synonyms, identifiers, functional group annotations, and physicochemical descriptors, and structurally similar molecules retrieved from the training set. We evaluate MolE-RAG across nine molecular property prediction tasks using proprietary, chemistry-specialized, and open-source LLMs. Across general-purpose LLMs, MolE-RAG improves ROC-AUC by up to 28 percentage points on classification tasks and reduces regression RMSE by up to 67% relative to a SMILES-only baseline. We further find that the utility of each context source varies across models and tasks, with different models benefiting most from textual retrieval, molecular context, or structural retrieval. These results suggest that molecule-centric retrieval can improve LLM-based molecular property prediction without model fine-tuning while providing a flexible framework for integrating heterogeneous chemical knowledge at inference time.

Problem

• molecular property prediction
• large language models
• chemical structure reasoning
• SMILES representation
• chemistry knowledge gap

Innovation

• Retrieval-Augmented Generation
• Molecular Property Prediction
• Large Language Models
• Chemical Knowledge Integration
• Structure-Enhanced Reasoning