One network to link them all: an untargeted multi-metabolomics-methods network
6 affiliations
- 1Metabolomics and Proteomics Core, Helmholtz Zentrum München, Ingostädter Landstraße 1, Neuherberg, Germany, 85764
- 2UMR1331 Toxalim
- 3Université de Toulouse, INRAE, ENVT, INP‑Purpan, UPS
- 4MetaboHUB
- 5National Infrastructure of Metabolomics and Fluxomics
- 6Chair of Analytical Food Chemistry, TUM School of Life Sciences, Technical University of Munich, Maximus-von-Imhof-Forum 2, Freising-Weihenstephan, Germany, 85354
Abstract
Metabolomics aims to cover a large range of metabolites from polar metabolites to hydrophobic lipids. Therefore, multiple chromatographic methods – such as reversed-phase liquid chromatography (RPLC) or hydrophilic interaction chromatography (HILIC) – are combined to provide a comprehensive overview. The resulting tables need to be combined afterward for data analysis, but several metabolites might be detected across multiple methods. Currently, it is only possible to deal with such multiple entries for identified metabolites. There is a strong requirement for data fusion strategies beyond the simple concatenation of individual peak tables. Networks have been shown to represent a powerful tool for metabolomics data analysis. Feature-based molecular networking, or ion identity networking, is a modern tool for reconstructing biochemical knowledge from metabolomics data. Here, we developed a molecular network across multiple acquisition methods to provide a more comprehensive combination of data and metabolite identification. MS/MS spectra between the different LC-method datasets were compared to merge features. Within single chromatographic methods, ion mode matching based on predefined adducts was used to identify metabolites. Within single-ionization modes, spectra were matched based on GNPS mass-shift cosine similarity. A large network of 4,862 nodes and 401,387 edges was generated based on the different connections. The network was tested on several blood-related matrices. Analyses were performed using RPLC-MS for mid- to non-polar metabolites and HILIC-MS for polar metabolites. Identification was performed using in-house and external spectral libraries, as well as prediction software such as Sirius. Well-known metabolites detected in all tested analytical methods, such as tryptophan, were successfully connected. Data were then mined for new metabolites and associations.
Related
- BEAR-GRN: Systematic assessment of single-cell multi-omics-based gene regulatory network inference methods.
- Inhibitory Mechanism of Bee Venom Against A431 Cutaneous Squamous Cell Carcinoma Cells Based on Label-Free Proteomics and Untargeted Metabolomics
- Short-Term Stable Catchment-Level Metabolic Structure in Urban Wastewater Revealed by Untargeted Metabolomics
- DNA Barcoding and Untargeted Metabolomics Reveal Phylogenetic and Chemotaxonomic Relationships Among 11 Ferula Species from Xinjiang, China
- Untargeted metabolomics in microbiome-immune crosstalk in cancer: challenges and recommendations for functional interpretation
BibTeX
@misc{couacault2026network,
title = {{One network to link them all: an untargeted multi-metabolomics-methods network}},
author = {Pauline Couacault and Sandra Costa González and Clément Frainay and Fabien Jourdan and Michael Witting},
howpublished = {ChemRxiv},
year = {2026},
month = sep,
doi = {10.26434/chemrxiv.15009370/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009370/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - One network to link them all: an untargeted multi-metabolomics-methods network
AU - Pauline Couacault
AU - Sandra Costa González
AU - Clément Frainay
AU - Fabien Jourdan
AU - Michael Witting
PY - 2026
DA - 2026/09/23
PB - ChemRxiv
DO - 10.26434/chemrxiv.15009370/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009370/v1
AB - Metabolomics aims to cover a large range of metabolites from polar metabolites to hydrophobic lipids. Therefore, multiple chromatographic methods – such as reversed-phase liquid chromatography (RPLC) or hydrophilic interaction chromatography (HILIC) – are combined to provide a comprehensive overview. The resulting tables need to be combined afterward for data analysis, but several metabolites might be detected across multiple methods. Currently, it is only possible to deal with such multiple entries for identified metabolites. There is a strong requirement for data fusion strategies beyond the simple concatenation of individual peak tables. Networks have been shown to represent a powerful tool for metabolomics data analysis. Feature-based molecular networking, or ion identity networking, is a modern tool for reconstructing biochemical knowledge from metabolomics data. Here, we developed a molecular network across multiple acquisition methods to provide a more comprehensive combination of data and metabolite identification. MS/MS spectra between the different LC-method datasets were compared to merge features. Within single chromatographic methods, ion mode matching based on predefined adducts was used to identify metabolites. Within single-ionization modes, spectra were matched based on GNPS mass-shift cosine similarity. A large network of 4,862 nodes and 401,387 edges was generated based on the different connections. The network was tested on several blood-related matrices. Analyses were performed using RPLC-MS for mid- to non-polar metabolites and HILIC-MS for polar metabolites. Identification was performed using in-house and external spectral libraries, as well as prediction software such as Sirius. Well-known metabolites detected in all tested analytical methods, such as tryptophan, were successfully connected. Data were then mined for new metabolites and associations.
ER -