Genetic Insights into Head-to-Body Ratios Via Deep Learning-Based Image Segmentation and Implications for Common Diseases

Head-to-body ratios (HBRs) are important anthropometric traits with direct relevance to human growth, development, and disease risk. However, the role of the proportions between head and body remains understudied, with the genetic basis of HBRs remaining largely unexplored. By applying deep learning models to 38,202 whole-body dual-energy X-ray absorptiometry images from the UK Biobank, we generated 10 distinct HBR phenotypes based on head (length/width) and various body dimensions. Our genome-wide association analyses identify 245 significant loci, with SNP-based heritability estimates ranging from 25% to 43%. Functional annotations show that genes prioritized for HBRs are enriched in chondrocytes in skeletal tissues and oligodendrocytes across multiple brain regions. Polygenic risk scores and mendelian randomization analyses further showed that HBRs are significantly associated with risks for cardiovascular, metabolic, musculoskeletal, and neuropsychiatric diseases, underscoring their potential value as health-related biomarkers. Evolutionary analyses show that HBR-associated variants are enriched in conserved genomic regions and human accelerated regions, particularly those influencing brain development. Overall, our study provides insights into the genetic architectures of HBRs, establishes their relevance to major human diseases, and offers evolutionary context for their biological significance.