Quality control and removal of technical variation of NMR metabolic biomarker data in ∼120,000 UK Biobank participants

Metabolic biomarker data quantified by nuclear magnetic resonance (NMR) spectroscopy for 249 circulating metabolites, lipids, and lipoprotein sub-fractions has recently become available in UK Biobank for approximately 121,657 participants. Here, we describe procedures for quality control and removal of technical variation for this biomarker data. We show that technical and biological effects with linear effects on individual biomarkers can combine in a non-linear fashion on (the 61) composite biomarkers and (81) biomarker ratios, and thus composite biomarkers and ratios should be re-derived after removal of unwanted variation to avoid introducing such cryptic effects. We make available an R package, ukbnmr, for extracting the metabolic biomarker data from UK Biobank and removing the unwanted technical variation. We also make available code for re-deriving the 61 composite biomarkers and 81 ratios, and for further derivation of 76 additional biomarker ratios of potential biological significance. Finally, we demonstrate that our removal of technical variation leads to increased signal for genetic and epidemiological studies of the NMR metabolic biomarkers in UK Biobank.