Effect heterogeneity reveals complex pleiotropic effects of rare coding variants

Genome-wide association studies (GWAS) and rare-variant association studies (RVAS) have identified thousands of genes and variants associated with diverse human phenotypes, and have revealed pervasive pleiotropy among common variants; however, the extent of pleiotropy among rare variants remains incompletely characterized. Using rare variant association results from the UK Biobank (UKB), we systematically quantify gene-level pleiotropy across hundreds of phenotypes spanning multiple phenotypic domains and identify genes exhibiting allelic series. We develop ALLSPICE (ALLelic Spectrum of Pleiotropy Informed Correlated Effects), a statistical test for assessing variant effect heterogeneity within pleiotropic genes across quantitative traits. In ALB, a gene associated with both albumin and calcium levels, ALLSPICE reveals a subset of missense variants with heterogeneous effects across these traits. Protein structure-based analyses further demonstrate that missense variants associated with calcium levels but not albumin, as well as variants with opposing effects on the two traits, are enriched near known calcium-binding sites in the ALB protein. In this way, we present a systematic view of gene-level pleiotropy and effect heterogeneity among rare variants, illustrating how shared and trait-specific genetic effects can arise within the same gene. Our analysis provides a general approach for dissecting pleiotropic architecture and detecting rare variants with heterogeneous effects for downstream biological investigation of complex human traits.