#32 · Genomics

Refining covariate adjustment and replicability assessment in GWAS and metabolomics

The dominant thread is correction of inference bias introduced by covariate adjustment in association studies. A Gaussian liability-model conversion from log odds is derived to correct collider bias in covariate-adjusted binary GWAS (pgen.1012348), and GEMRA adjusts phenotypes for medication use, with simulations showing less bias and higher power than common approaches (pgen.1012366). KDRD estimates the replicability of significant associations from a single study's summary statistics, tested on UK Biobank (biorxiv preprint). Two metabolomics resources extend this context: MetabHub mines UK Biobank plasma NMR metabolomics for mediation pathways, disease prediction and risk stratification, while MetabDB integrates 16 public metabolomics studies with QTLs, GWAS traits, diseases and genetic evidence. The new elements are methodological corrections applied at the phenotype and summary-statistic level, together with integrated resources that make metabolic association evidence searchable.