Genomics reveals eleven obesity endotypes with distinct biological and phenotypic signatures

Obesity, a leading global risk factor for cardiometabolic conditions, arises from multifaceted and biologically complex mechanisms1,2. To elucidate the full-dimensional genetic architecture underlying obesity, we conducted a multi-trait, multi-ancestry, genome-wide association study (GWAS) by combining genetic data on anthropometric traits (body mass index, waist circumference, waist-to-hip ratio, and hip circumference) from >2 million ancestrally diverse participants. We identified 743 significant loci, including 86 previously unreported loci, representing a 13% increase in locus discovery. We leveraged machine learning and multimodal data integration to identify the likely effector genes at obesity-associated loci. We performed genetic clustering on biologically enriched multi-trait GWAS data based on Bayesian non-negative matrix factorisation3 in the UK Biobank (n = 408,816) and identified 11 obesity clusters (endotypes). In addition to recapitulating the endotypes that drive classical metabolically healthy and unhealthy obesity phenotypes, we identified nine additional obesity endotypes driven by insulin physiology, beta cell compensation, immune dysregulation, neuroendocrine regulation, and lipid metabolism. Each cluster not only was characterised by unique biomarker features and clinical trajectories but also showed cluster-specific enrichment in tissue and single-cell regulatory regions. To facilitate the clinical and research adoption of obesity endotyping, we created partitioned polygenic scores for the 11 obesity endotypes, which we externally validated their performance using the Mass General Brigham Biobank (n = 48,377), and made publicly available in the PGS Catalog. This study marks a step forward in both the biological resolution and clinical translation of heterogeneity in obesity, implicating that obesity prevention and management should be as diverse as the condition itself.