henhaobio
Sign in

bioRxivSystems BiologyPreprint

MetaNetMap: automatic mapping of metabolomic data onto metabolic networks

Coralie MulleriD, Juliette AudemardiD, Sylvain PrigentiD, Clémence FriouxiD

PDFDOICodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

MotivationMetabolic networks represent genome-derived information about the biochemical reactions that cells are capable of performing. Mapping omic data onto these networks is important to refine model simulations. However, metabolomic data mapping remains very challenging due to difficulties in identifier reconciliation between annotation profiles and metabolic networks. ResultsMetaNetMap is a Python package designed to automatise the process of mapping metabolomic data onto metabolic networks. It includes several layers of identifier matching, the use of customisable databases, and molecular ontology integration to suggest the most matches between experimentally-identified metabolites and molecules defined in the network. We demonstrate its usability and the quality of automated mapping using two datasets. Availability and ImplementationMetaNetMap is an open source python package and publicly available under the GPLv3 licence. Source code is freely available on GitHub: https://github.com/coraliemuller/metanetmap. Data and code used in the application cases of this paper can be found at: https://doi.org/10.57745/ESFLR8.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 1 (5 Jan 2026) · CC BY 4.0

MetaNetMap automates mapping of metabolomic annotations onto genome-scale metabolic networks through direct, conversion-table-bridged and partial identifier matching, reproducing 311 mappings from a Parkinson's disease study.

Linking metabolites from metabolomic annotation profiles to genome-scale metabolic networks (GSMNs) is hindered by identifier mismatches, non-standard SMILES/InChI structures and partial annotations, so mapping is often manual and hard to reproduce. Building on cross-reference resources such as MetaNetX, MetaNetMap automates and logs the mapping while exploiting metadata from both input files.

Key results

  1. ShownOn a published Parkinson's disease (PD) metabolite list, automated mapping onto the human GSMN Recon3 3 produced 311 matches that agreed with the original study's mappings with identical outcomes, plus additional matches not reported in that study.

    With the MetaNetX vs MetaCyc reference tables, 344 vs 338 annotations mapped onto Recon3 3; 311 of these were also mapped in the original study with identical outcomes; 9 of the 311 were flagged as ambiguous. · Metabolomic mapping onto the human GSMN in the…; 4 Application; Tables

  2. ShownThe intermediary conversion table can recover matches that direct metadata comparison misses: six MetaNetX-only matches came from metabolomic InChI identifiers that did not directly match the GSMN and were bridged through the MetaNetX table to VMH identifiers.

    The 27 new matches found with the MetaCyc-based table were also found with MetaNetX, mostly via InChI and PubChem identifiers; six additional MetaNetX-only matches relied on InChI linked through MetaNetX to VMH identifiers in the GSMN. · Metabolomic mapping onto the human GSMN in the…

  3. SuggestedIn a mouse interventional metabolomics dataset with no prior mapping to compare against, automated mapping onto Recon3 3 reached coverage the authors consider typical for this task.

    Of 1019 annotated metabolites, 449 matched with MetaCyc (44.06%) and 442 with MetaNetX (43.37%); the authors cite an average rate of about 40% for mapping mass spectral standards to GSMNs. · Metabolomic mapping onto human GSMN in the…; Tables

  4. ShownOptional partial matching applied to unmatched entries, using ChEBI identifiers supplied in the mouse annotation profiles, raised mapping coverage for both reference databases.

    Coverage rose to 48.77% (MetaCyc) and 44.84% (MetaNetX) with partial matching; the authors describe these partial matches as curation opportunities for the metabolites concerned. · Metabolomic mapping onto human GSMN in the…; Partial matches; Tables

  5. SuggestedRemaining ambiguous matches trace to inconsistencies in the network's own metadata (PD case) and to formula-based matching (mouse case); the tool reports such multiple matches and leaves their resolution to the user.

    PD case: 9 ambiguous matches arose because some InChIs in Recon3 3 metadata point to two metabolites. Mouse case: some ambiguous matches remained, which the authors attribute to formula-based matching, since distinct metabolites can share a formula; these need manual checks. · Metabolomic mapping onto the human GSMN in the…; Metabolomic mapping onto human GSMN in the…; Handling Ambiguities

Methods, data and reproducibility

A Python package builds a unified conversion table from MetaCyc or MetaNetX files (plus optional custom identifiers). It then maps annotation profiles onto one GSMN (Classic mode) or several (Community mode) by direct SBML metadata matching, indirect table-bridged matching, and optional partial matching (ChEBI, InChIKey, enantiomer simplification).

  • Parkinson's disease metabolite list compiled from 87 human metabolomic studies, originally mapped onto Recon3 3 by the source study · human · n = 928 metabolites (87 studies)
  • Recon3 3, a version of the human metabolic model Recon3 3D, used as the target GSMN in both applications · human
  • Annotation profiles from a mouse study of myricetin cardioprotection after a week of high-intensity exercise, mapped onto Recon3 3 · mouse (mapped onto human GSMN Recon3 3) · n = 1019 metabolites
  • MetaCyc compounds.dat file used to build a reference conversion table
  • MetaNetX chem_xref and chem_prop files used to build a reference conversion table
  • Data and code for the application cases · 10.57745/ESFLR8
Methods & tools
MetaNetMap v1.1.0, MetaNetX, MetaCyc, Recon3 3 (human GSMN), SBML, ChEBI, InChI / InChIKey, PubChem, VMH identifiers, SMILES
Software & environment
MetaNetMap v1.1.0 (open-source Python package, GPLv3 licence) used for the applications; no other dependency versions or containers stated.
Assay conditions & compute
Mapping time is reported in Table 1 for a personal laptop with 64Gb RAM; the authors call it relatively low without partial matching and note that MetaCyc's smaller size relative to MetaNetX strongly affects performance. Timing values are not given in the text.
Reproducibility gaps
Preprocessing, reference database construction, partial matching strategies and detailed application methods are deferred to Supplementary material not included here; Table 1 values are not reproduced in the text.
Limitations (6)
  • Ambiguous multi-matches (one-to-many or many-to-one between profiles, conversion table and GSMN) are reported but not resolved; resolving them is left to the user, and some require manual verification. · stated by the authors
  • Matching on molecular formulas raises coverage but causes ambiguity, because distinct metabolites can share a formula; the authors advise against it unless no other match exists. · stated by the authors
  • Partial matching could not be applied to the Parkinson's disease dataset because its metabolomic data lacked the additional metadata needed. · stated by the authors
  • The mouse dataset had no original mapping, so its mapping results could not be compared directly against a reference. · stated by the authors
  • Validation rests on two case studies, both mapped onto the same human GSMN, and only one has a reference mapping. Precision against a curated gold standard is not reported, and no other mapping tool is benchmarked. · noted in this summary
  • The correctness of partial matches is not quantified, so the coverage gains from ChEBI-based partial matching may include false-positive assignments. · noted in this summary

Read this if you need to map metabolomic annotation profiles onto one or several genome-scale metabolic networks and want an automated, logged alternative to manual identifier reconciliation.

Citations
1

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISRahnavard '18Schaul '26Köhler '23Mandwal '23Reiman '21Shen '19
64% semantic matchBioinformatics
Ali Rahnavard, Daniel S. Hitchcock, Julián Ávila-Pacheco+7 more·bioRxiv·Oct 16, 2018· 5 citations