Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Comparative Analysis of Formula and Structure Prediction from Tandem Mass Spectra

Xujun Che, Xiuxia Du, Depeng Xu

TL;DR

This study systematically evaluates state-of-the-art formula and structure prediction methods for tandem mass spectra in LC-MS-based metabolomics and exposomics, using NIST 23 and MoNA datasets. It contrasts SIRIUS and MIST-CF for formula prediction and DiffMS and MSNovelist for structure generation, examining performance across adduct types and end-to-end pipelines with in-domain retraining. The results show practical accuracy for dominant adducts but reveal weaknesses for water-loss and rare adducts, highlighting that structure generation is the primary bottleneck and that domain adaptation greatly boosts performance. The work provides actionable guidance on evaluation practices, data curation, and future directions such as multimodal integration and specialized architectures to handle underrepresented adducts.

Abstract

Liquid chromatography mass spectrometry (LC-MS)-based metabolomics and exposomics aim to measure detectable small molecules in biological samples. The results facilitate hypothesis-generating discovery of metabolic changes and disease mechanisms and provide information about environmental exposures and their effects on human health. Metabolomics and exposomics are made possible by the high resolving power of LC and high mass measurement accuracy of MS. However, a majority of the signals from such studies still cannot be identified or annotated using conventional library searching because existing spectral libraries are far from covering the vast chemical space captured by LC-MS/MS. To address this challenge and unleash the full potential of metabolomics and exposomics, a number of computational approaches have been developed to predict compounds based on tandem mass spectra. Published assessment of these approaches used different datasets and evaluation. To select prediction workflows for practical applications and identify areas for further improvements, we have carried out a systematic evaluation of the state-of-the-art prediction algorithms. Specifically, the accuracy of formula prediction and structure prediction was evaluated for different types of adducts. The resulting findings have established realistic performance baselines, identified critical bottlenecks, and provided guidance to further improve compound predictions based on MS.

Comparative Analysis of Formula and Structure Prediction from Tandem Mass Spectra

TL;DR

This study systematically evaluates state-of-the-art formula and structure prediction methods for tandem mass spectra in LC-MS-based metabolomics and exposomics, using NIST 23 and MoNA datasets. It contrasts SIRIUS and MIST-CF for formula prediction and DiffMS and MSNovelist for structure generation, examining performance across adduct types and end-to-end pipelines with in-domain retraining. The results show practical accuracy for dominant adducts but reveal weaknesses for water-loss and rare adducts, highlighting that structure generation is the primary bottleneck and that domain adaptation greatly boosts performance. The work provides actionable guidance on evaluation practices, data curation, and future directions such as multimodal integration and specialized architectures to handle underrepresented adducts.

Abstract

Liquid chromatography mass spectrometry (LC-MS)-based metabolomics and exposomics aim to measure detectable small molecules in biological samples. The results facilitate hypothesis-generating discovery of metabolic changes and disease mechanisms and provide information about environmental exposures and their effects on human health. Metabolomics and exposomics are made possible by the high resolving power of LC and high mass measurement accuracy of MS. However, a majority of the signals from such studies still cannot be identified or annotated using conventional library searching because existing spectral libraries are far from covering the vast chemical space captured by LC-MS/MS. To address this challenge and unleash the full potential of metabolomics and exposomics, a number of computational approaches have been developed to predict compounds based on tandem mass spectra. Published assessment of these approaches used different datasets and evaluation. To select prediction workflows for practical applications and identify areas for further improvements, we have carried out a systematic evaluation of the state-of-the-art prediction algorithms. Specifically, the accuracy of formula prediction and structure prediction was evaluated for different types of adducts. The resulting findings have established realistic performance baselines, identified critical bottlenecks, and provided guidance to further improve compound predictions based on MS.
Paper Structure (29 sections, 1 figure, 6 tables)

This paper contains 29 sections, 1 figure, 6 tables.

Table of Contents

  1. Introduction
  2. Experimental Section
  3. Datasets
  4. NIST 23 Spectral Library:
  5. MoNA (MassBank of North America:)
  6. Dataset Statistics:
  7. Computational Methods for Compound Prediction
  8. Chemical Formula Prediction
  9. SIRIUS:
  10. MIST-CF:
  11. Molecular Structure Generation
  12. MSNovelist:
  13. DiffMS:
  14. Evaluation Setup
  15. Metrics:
  16. ...and 14 more sections

Figures (1)

  • Figure 1: Diagnostic analysis of pipeline performance. (a) Structure generation robustness to formula errors. (b) Failure mode decomposition: structure generation bottleneck (b1), formula propagation error (b2), and rare adduct collapse (b3). (c) Isomer discrimination challenge with 66.7% success rate for C$_{20}$H$_{32}$O$_3$.