FraGNNet: A Deep Probabilistic Model for Mass Spectrum Prediction

2024-04-02

arXiv.org

AI Summary

Existing compound-to-tandem mass spectrometry (C2MS) prediction models suffer from low mass accuracy, poor generalizability, limited interpretability, and scalability challenges. To address these issues, this paper proposes a deep probabilistic model integrating graph neural networks with variational inference. Our method introduces a structured latent space to explicitly model the MS/MS generation mechanism: it jointly encodes molecular graph structure, explicitly represents fragmentation pathways, and probabilistically estimates both peak intensities and their uncertainties. The resulting framework enables end-to-end, high-resolution, and interpretable spectrum generation. Experimentally, it achieves state-of-the-art prediction accuracy—measured by spectral error metrics—and significantly outperforms all baseline methods in retrieval-based compound identification. Notably, it markedly improves identification rates for low-abundance and uncharacterized compounds in complex mixtures, demonstrating superior robustness and practical utility in real-world metabolomics applications.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

The process of identifying a compound from its mass spectrum is a critical step in the analysis of complex mixtures. Typical solutions for the mass spectrum to compound (MS2C) problem involve matching the unknown spectrum against a library of known spectrum-molecule pairs, an approach that is limited by incomplete library coverage. Compound to mass spectrum (C2MS) models can improve retrieval rates by augmenting real libraries with predicted spectra. Unfortunately, many existing C2MS models suffer from problems with prediction resolution, scalability, or interpretability. We develop a new probabilistic method for C2MS prediction, FraGNNet, that can efficiently and accurately predict high-resolution spectra. FraGNNet uses a structured latent space to provide insight into the underlying processes that define the spectrum. Our model achieves state-of-the-art performance in terms of prediction error, and surpasses existing C2MS models as a tool for retrieval-based MS2C.

Problem

• Predicting tandem mass spectra from molecular structures accurately
• Overcoming limitations of incomplete spectral libraries in compound identification
• Improving mass accuracy and generalization in spectrum prediction models

Innovation

• Deep probabilistic model for MS/MS prediction
• Learns distribution over molecule fragments
• Achieves high mass accuracy and generalization