Integration of metabolomic and genetic data reveals novel variants underpinning the human metabolome: the Coronary Artery Risk Development in Young Adults (CARDIA) study

Metabolite genome-wide association studies have identified hundreds of variants, many of which play intermediate roles linking genotype and phenotype with downstream diseases. However, the majority of metabolite GWAS have been published in self-identified Non-Hispanic White (NHW) populations, greatly limiting inference to other populations. Here we report the results of a GWAS of 7,522 untargeted metabolite peaks in 2,183 participants of the Coronary Artery Risk Development in Young Adults (CARDIA) study (714 Black and 1,469 White individuals, mean age = 43.56, women = 56.1%). We performed untargeted metabolomic profiling on plasma samples from these individuals using ultra-high performance liquid chromatography-high resolution mass spectrometry. Using race/ethnicity-stratified and combined GWAS of 8,534,915 (Black stratum) and 5,886,255 (White stratum) common variants (minor allele frequency ≥ 0.05) from TOPMed Imputation Reference Panel r2, we examined association with the CARDIA 7,522 metabolite peaks. We used MetaboAnalyst 5.0 and publicly available phenome-wide association studies to predict functional pathway activity and characterize unknown peaks. We identified 171 significant (P-value -13) sentinel variants across 536 metabolite peaks, of which 39 were annotated with our in-house library in the race-combined CARDIA analysis. PheWAS and pathway enrichment models provided information on unannotated metabolite peaks including the novel KHNYN/SDR39U1 locus which supports vitamin B5's involvement in the citric acid cycle. We also identified a race-specific variant (rs79530723) near SLC28A1, which plays an important role in cytidine metabolism and was specific to the CARDIA Black participants. Pathway analyses prioritized a role of pyrimidine metabolism for this peak, a finding supported by studies in the Hispanic/Latino populations. Overall, our study results support a strong shared genetic architecture of the human metabolome across diverse populations and reveal new metabolite GWAS signals, underscoring the value of integrating 'omics' techniques to enhance comprehension of its genetic architecture.