Abstract
Gene discoveries in obesity have largely relied on homogeneous populations, limiting their generalizability across ancestries. Here, we conduct a gene-based rare variant association study of BMI on 839,110 individuals from six ancestries across two population-scale biobanks. A cross-ancestry meta-analysis identifies 13 genes, including YLPM1, RIF1, GIGYF1, SLC5A3, and GRM7, that confer about three-fold risk for severe obesity, are expressed in the brain and adipose tissue, and are linked to obesity traits such as body-fat percentage. While YLPM1, MC4R, and SLTM show consistent effects, GRM7 and APBA1 show significant ancestral heterogeneity. Polygenic risk additively increases obesity penetrance, and phenome-wide studies reveal additional associations, including YLPM1 with altered mental status. These genes also influence cardiometabolic comorbidities, including GIGYF1 and SLTM towards type 2 diabetes with or without BMI as a mediator, and altered levels of plasma proteins, such as LECT2 and NCAN, which in turn affect BMI. Our findings provide insights into the genetic basis of obesity and its related comorbidities across ancestries and ascertainments.