Non-targeted metabolite profiling using ultra performance liquid chromatography-mass spectrometry (UPLCMS)
was performed as part of a large-scale epidemiological study involving biobanked serum samples. The influence of
both biological (age and body mass index) and technical (season of sample collection, fasting time, handling time, and
storage time) covariates on the analysis was assessed. Statistical models including different sets of these covariates were
compared and the results illustrate that variation in which covariates were included did not have an appreciable effect on
the number or composition of biologically significant metabolite features associated with body mass index or age.
Furthermore, when all covariates were included in the model, there was little overlap of metabolite features significantly
associated with the different covariates. Thus, the results of this study illustrate that while some of the observed quantitative
variance of metabolite features can be explained by biological and technical covariates, the use of non-targeted
metabolite profiling of serum by UPLC-MS is valid for studies of biological outcomes in biobanked clinical samples from
Keywords: Biobanked clinical samples, biomarker discovery, covariates, Non-targeted metabolomics, statistical models, ultra
performance liquid chromatography-mass spectrometry.
open access plus
Rights & PermissionsPrintExport