Environment and diet in early life can affect development and health throughout the life course. Metabolic phenotyping of urine and serum represents a complementary systems-wide approach to elucidate environment–health interactions. However, large-scale metabolome studies in children combining analyses of these biological fluids are lacking. Here, we sought to characterise the major determinants of the child metabolome and to define metabolite associations with age, sex, BMI and dietary habits in European children, by exploiting a unique biobank established as part of the Human Early-Life Exposome project ( http://www.projecthelix.eu ). Metabolic phenotypes of matched urine and serum samples from 1192 children (aged 6–11) recruited from birth cohorts in six European countries were measured using high-throughput 1H nuclear magnetic resonance (NMR) spectroscopy and a targeted LC-MS/MS metabolomic assay (Biocrates AbsoluteIDQ p180 kit). We identified both urinary and serum creatinine to be positively associated with age. Metabolic associations to BMI z-score included a novel association with urinary 4-deoxyerythronic acid in addition to valine, serum carnitine, short-chain acylcarnitines (C3, C5), glutamate, BCAAs, lysophosphatidylcholines (lysoPC a C14:0, lysoPC a C16:1, lysoPC a C18:1, lysoPC a C18:2) and sphingolipids (SM C16:0, SM C16:1, SM C18:1). Dietary-metabolite associations included urinary creatine and serum phosphatidylcholines (4) with meat intake, serum phosphatidylcholines (12) with fish, urinary hippurate with vegetables, and urinary proline betaine and hippurate with fruit intake. Population-specific variance (age, sex, BMI, ethnicity, dietary and country of origin) was better captured in the serum than in the urine profile; these factors explained a median of 9.0% variance amongst serum metabolites versus a median of 5.1% amongst urinary metabolites. Metabolic pathway correlations were identified, and concentrations of corresponding metabolites were significantly correlated (r > 0.18) between urine and serum. We have established a pan-European reference metabolome for urine and serum of healthy children and gathered critical resources not previously available for future investigations into the influence of the metabolome on child health. The six European cohort populations studied share common metabolic associations with age, sex, BMI z-score and main dietary habits. Furthermore, we have identified a novel metabolic association between threonine catabolism and BMI of children.

Untargeted metabolomics (including lipidomics) is a holistic approach to biomarker discovery and mechanistic insights into disease onset and progression, and response to intervention. Each step of the analytical and statistical pipeline is crucial for the generation of high-quality, robust data. Metabolite identification remains the bottleneck in these studies; therefore, confidence in the data produced is paramount in order to maximize the biological output. Here, we outline the key steps of the metabolomics workflow and provide details on important parameters and considerations. Studies should be designed carefully to ensure appropriate statistical power and adequate controls. Subsequent sample handling and preparation should avoid the introduction of bias, which can significantly affect downstream data interpretation. It is not possible to cover the entire metabolome with a single platform; therefore, the analytical platform should reflect the biological sample under investigation and the question(s) under consideration. The large, complex datasets produced need to be pre-processed in order to extract meaningful information. Finally, the most time-consuming steps are metabolite identification, as well as metabolic pathway and network analysis. Here we discuss some widely used tools and the pitfalls of each step of the workflow, with the ultimate aim of guiding the reader towards the most efficient pipeline for their metabolomics studies.