Can we trust untargeted metabolomics? Results of the metabo-ring initiative, a large-scale, multi-instrument inter-laboratory study

Jean-Charles Martin, Matthieu Maillot, Gérard Mazerolles, Alexandre Verdu, Bernard Lyan, Carole Migné, Catherine Defoort, Cecile Canlet, Christophe Junot, Claude Guillou, Claudine Manach, Daniel Jabob, Delphine Jouan-Rimbaud Bouveresse, Estelle Paris, Estelle Pujos-Guillot, Fabien Jourdan, Franck Giacomoni, Frédérique Courant, Gaëlle Favé, Gwenaëlle Le GallHubert Chassaigne, Jean-Claude Tabet, Jean-Francois Martin, Jean-Philippe Antignac, Laetitia Shintu, Marianne Defernez, Mark Philo, Marie-Cécile Alexandre-Gouaubau, Marie-Josephe Amiot-Carlin, Mathilde Bossis, Mohamed N. Triba, Natali Stojilkovic, Nathalie Banzet, Roland Molinié, Romain Bott, Sophie Goulitquer, Stefano Caldarelli, Douglas N. Rutledge

Research output: Contribution to journalArticlepeer-review

74 Citations (Scopus)

Abstract

The metabo-ring initiative brought together five nuclear magnetic resonance instruments (NMR) and 11 different mass spectrometers with the objective of assessing the reliability of untargeted metabolomics approaches in obtaining comparable metabolomics profiles. This was estimated by measuring the proportion of common spectral information extracted from the different LCMS and NMR platforms. Biological samples obtained from 2 different conditions were analysed by the partners using their own in-house protocols. Test #1 examined urine samples from adult volunteers either spiked or not spiked with 32 metabolite standards. Test #2 involved a low biological contrast situation comparing the plasma of rats fed a diet either supplemented or not with vitamin D. The spectral information from each instrument was assembled into separate statistical blocks. Correlations between blocks (e.g., instruments) were examined (RV coefficients) along with the structure of the common spectral information (common components and specific weights analysis). In addition, in Test #1, an outlier individual was blindly introduced, and its identification by the various platforms was evaluated. Despite large differences in the number of spectral features produced after post-processing and the heterogeneity of the analytical conditions and the data treatment, the spectral information both within (NMR and LCMS) and across methods (NMR vs. LCMS) was highly convergent (from 64 to 91 % on average). No effect of the LCMS instrumentation (TOF, QTOF, LTQ-Orbitrap) was noted. The outlier individual was best detected and characterised by LCMS instruments. In conclusion, untargeted metabolomics analyses report consistent information within and across instruments of various technologies, even without prior standardisation.

Original languageEnglish
Pages (from-to)807-821
Number of pages15
JournalMetabolomics
Volume11
Issue number4
Early online date14 Oct 2014
DOIs
Publication statusPublished - Aug 2015
Externally publishedYes

Keywords

  • Inter-laboratory
  • Mass spectrometry
  • Metabolic fingerprinting
  • Nuclear magnetic resonance
  • Untargeted metabolomics

Cite this