Metabolites refer to any small molecules involved in metabolism as an intermediate or end product. Human metabolite concentrations can be assayed in blood, or less intrusively in urine. Urinary metabolites can reflect the function of biological pathways, including energy metabolism, diet, and amino-acid metabolism [1]. Metabolite concentrations have been implicated as biomarkers for physical and psychological health outcomes ranging from kidney disease to psychiatric disorders [2-6].
The concentrations of some metabolites are highly heritable, with previous research identifying hundreds of metabolite-gene associations [2, 7, 8]. Traits have distinct distributions of SNP effect sizes (genetic architectures) which influence the choice and performance of statistical methods. For example, methods such as LDSC regression assume a highly polygenic genetic architecture and may not be appropriate for traits with a more oligogenic architecture.
SbayesR is a statistical tool that uses Bayesian posterior inference and GWAS summary data to model trait architecture while simultaneously estimating SNP-based heritability [9]. We estimated the genetic architecture of 53 urinary metabolite concentrations in SbayesR using summary statistics from the current largest GWAS of urinary metabolites by Valo et al. (2025) [2].
We detected substantial variation in genetic architectures across the 53 urinary metabolites. Of the 29 metabolite concentrations with a detectable SNP-based heritability, 12 showed architectures dominated by fewer, large-effect SNPs. The remaining metabolites displayed more evenly spread (polygenic) SNP-effect distributions. Our SNP-based heritability estimates were generally lower than that of Valo et al.’s meta-analysis (except Taurine), likely reflecting the use of summary data as opposed to primary genotype data. Our posterior parameter estimates showed little sensitivity to model priors.
These results are informative for future genetic analyses of urinary metabolomic data and support previous evidence for large-effect loci for several metabolites [2].