Can Metabolomic Profiles Reproduce Sample Differences Across Laboratories?
An eight-laboratory US and UK round-robin study found that metabolomic profiles could differentiate known heterogeneous sample mixtures across laboratories using their own LC-MS protocols. The study also identified factors that affect direct comparison of metabolite measurements between laboratories.
Study at a Glance
| Journal | Analytical Chemistry, 2026 |
|---|---|
| Laboratories | Eight, in the US and UK, each using its own LC-MS methods |
| Samples | 32 samples: duplicate sets of 16 defined beef, chicken and pork extract mixtures |
| Features detected | From 560 to 1,727 per laboratory |
| Mixture differentiation | Adjusted Rand index of 1 in seven laboratories and 0.84 in the eighth (non-negative matrix factorization) |
| Inter-laboratory agreement | Mean Pearson correlation of 0.87 ± 0.12 for suppression-corrected abundances of metabolites matched with authentic standards |
What the researchers tested
Each laboratory received identical aliquots of 32 samples comprising duplicate sets of 16 defined beef, chicken, and pork extract mixtures and concentrations. Samples were prepared centrally, analyzed with each laboratory’s methods, and processed and evaluated centrally using the TruQuant approach.
What the study found
Feature detection varied from 560 to 1,727 across laboratories, yet all laboratories differentiated the sample mixture types using unsupervised approaches. For non-negative matrix factorization, the adjusted Rand index was 1 in seven laboratories and 0.84 in the remaining laboratory.
Suppression-corrected abundances of metabolites matched using authentic standards showed a mean inter-laboratory Pearson correlation of 0.87 ± 0.12. This correlation reflects agreement in abundance patterns across samples; it does not establish identical concentrations between laboratories.
Choosing the right data output
Normalized data supported separation of sample groups by reducing concentration-related variation. However, normalization introduced some laboratory-specific signal dependence. Suppression-corrected data were better suited to direct inter-laboratory feature-abundance comparisons in this study. These outputs therefore serve different analytical purposes.
Scope of the evidence
The round-robin study demonstrates that using the same internal standard across different labs with different SOPs and instruments, all labs got the same answer: they correctly identified the same sample compositions and the same biomarkers. It is clear that the ability to correct instrumental and sample-to-sample variances to a common internal standard is a mechanism to produce accurate and reproducible results.
Metabolites in this study were matched using authentic standards. IROA’s authentic metabolite libraries of standards support the same kind of confident matching in your own laboratory.