Explainable Molecular Structure Inference from GC--MS with Diffusion Models and LLM Reranking

2026-10-02

AI Summary

This study addresses the limitations of traditional GC-MS analysis, which relies heavily on spectral library matching and struggles to identify unknown compounds or directly infer molecular structures. To overcome these challenges, this work proposes DiffGCMS, a novel framework that introduces a two-stage "generation plus LLM reasoning" architecture for de novo interpretation of mass spectra into molecular structures. Specifically, the method employs a discrete graph diffusion model to generate candidate structures, followed by a large language model that performs verification, reranking, and fragment ion evidence tracing, thereby enabling library-free compound identification. Experimental evaluations on the NIST test set demonstrate that DiffGCMS achieves an Acc@1 of 6.01%, which improves to 21.95% on a small-molecule subset after LLM-based refinement, with candidate validity reaching 100%.

Abstract

GC--EI--MS is an important technique for analyzing volatile and semivolatile compounds in complex samples. However, conventional methods rely heavily on reference spectral library matching, limiting their ability to identify compounds absent from these libraries and to infer complete molecular structures directly from fragmentation information. Here, we present DiffGCMS, a spectrum-conditioned discrete graph diffusion model for de novo structure elucidation from GC--EI--MS, and further develop a framework that integrates DiffGCMS with second-stage reasoning by a large language model (LLM). In the first stage, DiffGCMS generates candidate molecular structures from input spectra; in the second stage, the LLM uses mass spectral information to validate, repair, and rerank the candidates and provides interpretable analysis of fragment-ion peaks. This framework can generate plausible molecular structures for compounds absent from reference spectral libraries and provide traceable evidence supporting its decisions. On a test set comprising 13,696 spectra from NIST 20, the generative model achieved Acc@1 and Acc@10 of 6.01\% and 15.76\%, respectively. On the test subset containing molecules with no more than 10 heavy atoms, LLM-assisted molecular graph repair and reranking increased Acc@1 from 21.28\% to 21.95\%, Acc@10 from 46.91\% to 47.99\%, and candidate validity from 91.04\% to 100\%. These results demonstrate that spectrum-aware postprocessing can correct errors produced by the generative model while providing auditable and traceable explanations for the final ranking.

Problem

• GC-EI-MS
• molecular structure inference
• de novo structure elucidation
• spectral library matching
• explainability

Innovation

• Discrete Graph Diffusion Model
• Large Language Model
• GC-EI-MS
• De Novo Structure Elucidation
• Explainable AI