The phenomic landscape of rare gain- and loss-of-function variation

Current large-scale phenome-wide association studies do not distinguish between genetic variants with distinct functional consequences, despite evidence that gain-of-function (GoF) and loss-of-function (LoF) variants can produce different phenotypes. Consequently, the phenotypic consequences of variants stratified by predicted GoF and LoF effects have not been systematically characterized at scale. Here, we systematically analyzed rare predicted GoF and LoF variants in 18,214 genes across 1,244 binary and 202 quantitative traits using whole-exome sequencing and electronic health records from the Mount Sinai Million Health Discoveries Program, a phenotypically rich hospital-based biobank comprised of 50,067 participants, primarily representing African, Admixed American and European ancestries. We identified numerous established and previously unreported ancestry-specific and shared gene- and variant-level associations, many of which were supported by complementary evidence from both diagnostic phenotypes and clinical measurements. Proteomic evaluation of the previously unreported association between VNN1 LoF variants and body mass index in the Multi-Ethnic Study of Atherosclerosis revealed reduced circulating vanin-1 levels with LoF variants, which in turn were associated with body composition and additional cardiometabolic phenotypes. Functional stratification of predicted GoF and LoF variants revealed gene-trait associations that were attenuated when these variants were combined, demonstrating that traditional aggregation of variants with different functional effects can obscure association signals. These findings underscore the importance of incorporating predicted functional consequences into association analyses and establish a comprehensive resource linking exome-wide predicted GoF and LoF variants to the human phenome across ancestry-specific and multi-ancestry analyses.