Osmolality-based normalization enhances statistical discrimination of untargeted metabolomic urine analysis: results from a comparative study

• Published in: Metabolomics Vol. 17; no. 1; p. 2
• Main Authors: Mervant, Loïc, Tremblay-Franco, Marie, Jamin, Emilien L., Kesse-Guyot, Emmanuelle, Galan, Pilar, Martin, Jean-François, Guéraud, Françoise, Debrauwer, Laurent
• Format: Journal Article
• Language: English
• Published: New York Springer US 2021; Springer Nature B.V; Springer Verlag
• Subjects: Biochemistry; Biomedical and Life Sciences; Biomedicine; Cell Biology; Developmental Biology; Life Sciences; Liquid chromatography; Mass spectroscopy; Mathematical models; Metabolomics; Methods; Molecular Medicine; Original Article; Statistical analysis; Toxicology; Urine; Normalization; Mass spectrometry; Osmolality; Untargeted metabolomics; Urine analysis
• ISSN: 1573-3882; 1573-3890
• DOI: 10.1007/s11306-020-01758-z

Introduction
Because of its ease of collection, urine is one of the most commonly used matrices for metabolomics studies. However, unlike other biofluids, urine exhibits tremendous variability that can introduce confounding inconsistency during result interpretation. Despite many existing techniques to normalize urine samples, there is still no consensus on either which method is most appropriate or how to evaluate these methods.
Objectives
To investigate the impact of several methods and combinations of methods conventionally used in urine metabolomics on the statistical discrimination of two groups in a simple metabolomics study.
Methods
We applied 14 different strategies of normalization to forty urine samples analysed by liquid chromatography coupled to high-resolution mass spectrometry (LC-HRMS). To evaluate the impact of these different strategies, we relied on the ability of each method to reduce confounding variability while retaining variability of interest, as well as the predictability of statistical models.
Results
Among all tested normalization methods, osmolality-based normalization gave the best results. Moreover, we demonstrated that normalization using a specific dilution prior to the analysis outperformed post-acquisition normalization. We also demonstrated that the combination of various normalization methods does not necessarily improve statistical discrimination.
Conclusions
This study re-emphasized the importance of normalizing urine samples for metabolomics studies. In addition, it appeared that the choice of method had a significant impact on result quality. Consequently, we suggest osmolality-based normalization as the best method for normalizing urine samples.
Trial registration number: NCT03335644