Validation of an LC-HRMS-based untargeted metabolomics analytical platform for a long-term clinical metabolomics study

Untargeted metabolomics by LC-HRMS is a powerful tool when searching for new biomarkers to enhance our knowledge of pathophysiological processes. When used for large-scale clinical studies, it is crucial to have a comprehensive and robust setup and that the laboratory’s ability to collect reliable data is documented. This study aimed to establish a robust analytical platform for a long-term clinical metabolomics study, assessing the method's fitness-for-purpose. Samples were prepared with an automated liquid handler and four analytical methods were developed and evaluated. The method validation study spanned three batches with twelve runs using individual serum samples and quality control samples. Data was acquired with untargeted acquisition and only metabolites identified with a local mass spectral library including retention time (level 1 identification) were evaluated. Validation parameters included reproducibility, repeatability, stability, and identification selectivity, emphasizing dataset intrinsic variance. Of the four methods tested, two were selected for validation. A total of 47 metabolites in RPLC-ESI+-and 55 metabolites in HILIC-ESI--HRMS met specified validation criteria, and full validation reports are presented. Quality assurance involved system suitability testing, sample release, run release, and batch release. Concentrated and diluted pooled quality control samples were introduced to correct for variance in response linearity across batches. The acquired data files can further be revisited for ad hoc validation of future metabolome and exposome features detected with the same setup. The validated methods proved fit-for-purpose. The laboratory demonstrated its capability to produce reliable results for large-scale, clinical, untargeted metabolomics. This study underscores the value of validation of an assays performance before commencing larger studies to enhance the reliability and impact of the generated hypotheses.