Diffusion models and language AI improve molecular structure identification
Explainable Molecular Structure Inference from GC--MS with Diffusion Models and LLM Reranking
Machine Learning
Summary
Analyzing chemicals by their mass patterns is hard, especially if the chemical isn’t in known databases. The authors developed a two-step method: first, a model guesses possible molecular shapes from data patterns; then, a language AI checks and fixes these guesses, also explaining how it made decisions. This approach helps find new molecules and provides clear reasons for its results. It tested well on a large set of chemical data.
What this means in practice
- •For chemical analysis labs: Identify unknown volatile compounds outside existing spectral libraries by combining AI-generated structures and language model validation.
- •For pharmaceutical development teams: Improve early-stage compound identification by generating and validating possible molecular structures directly from mass spectrometry data.
Authors
Changlin Liu, Tianyu Yi, Chengchun Liu, Boxuan Zhao, Fanyang Mo
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.