De novo molecular structure elucidation from mass spectra via flow matching

2026-02-23

AI Summary

Directly reconstructing complete small-molecule structures from mass spectrometry (MS) data is an ill-posed inverse problem, and conventional approaches suffer from limited accuracy. This work proposes MSFlow, the first two-stage framework to incorporate flow matching generative models into MS interpretation. In the first stage, a molecular formula-constrained Transformer maps the input spectrum into a continuous embedding rich in chemical information. The second stage employs a discrete flow matching decoder to generate valid molecular structures from this embedding. By integrating information-preserving molecular descriptors with a structure-aware generative mechanism, MSFlow achieves a Top-1 accuracy of 45% on small-molecule structure elucidation—up to 14 times higher than the current state-of-the-art methods.

Technology Category

• Machine Learning: Structured Learning
• Planning, Routing, and Scheduling: Planning with Language Models
• Natural Language Processing: Generation

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Bridging structured and unstructured data
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Mass spectrometry is a powerful and widely used tool for identifying molecular structures due to its sensitivity and ability to profile complex samples. However, translating spectra into full molecular structures is a difficult, under-defined inverse problem. Overcoming this problem is crucial for enabling biological insight, discovering new metabolites, and advancing chemical research across multiple fields. To this end, we develop MSFlow, a two-stage encoder-decoder flow-matching generative model that achieves state-of-the-art performance on the structure elucidation task for small molecules. In the first stage, we adopt a formula-restricted transformer model for encoding mass spectra into a continuous and chemically informative embedding space, while in the second stage, we train a decoder flow matching model to reconstruct molecules from latent embeddings of mass spectra. We present ablation studies demonstrating the importance of using information-preserving molecular descriptors for encoding mass spectra and motivate the use of our discrete flow-based decoder. Our rigorous evaluation demonstrates that MSFlow can accurately translate up to 45 percent of molecular mass spectra into their corresponding molecular representations - an improvement of up to fourteen-fold over the current state-of-the-art. A trained version of MSFlow is made publicly available on GitHub for non-commercial users.

Problem

• molecular structure elucidation
• mass spectrometry
• inverse problem
• de novo

Innovation

• flow matching
• mass spectrometry
• molecular structure elucidation
• generative model
• transformer