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

Description of the data and file structure By applying deep learning models to 38,202 whole-body dual-energy X-ray absorptiometry (DXA) images from the UK Biobank, we extracted 10 HBRs related to the head (length and width) and body (height, shoulder width, trunk length, hip width, and leg length). Files and variables File: 1.LHiR.stat.gz Description: GWAS summary data of head length to hip width ratio. File: 1.LHR.stat.gz Description: GWAS summary data of head length to body height ratio. File: 1.LSR.stat.gz Description: GWAS summary data of head length to shoulder width ratio. File: 1.LLeR.stat.gz Description: GWAS summary data of head length to leg length ratio. File: 1.LTR.stat.gz Description: GWAS summary data of head length to trunk length ratio. File: 1.WHiR.stat.gz Description: GWAS summary data of head width to hip width ratio. File: 1.WHR.stat.gz Description: GWAS summary data of head width to body height ratio. File: 1.WLeR.stat.gz Description: GWAS summary data of head width to leg length ratio. File: 1.WSR.stat.gz Description: GWAS summary data of head width to shoulder width ratio. File: 1.WTR.stat.gz Description: GWAS summary data of head width to trunk length ratio. Code/software GWASs were performed using BOLT-LMM. Covariates included the first 20 genetic principal components (FID 22009) provided by UKB, sex (FID 31), age (FID 21003) and "BACKGROUND". Additionally, the DXA scanner's serial number and the software version used to process images were combined into a single covariate, resulting in seven factor levels. The codes used for quality control of the DXA images and for performing deep learning-based segmentation of the head and key landmarks are available.