Identification of genetic loci associated with aortic stiffness using deep learning and genome-wide association analysis on 45,789 UK Biobank participants

Abstract Background Aortic stiffness is a key predictor of cardiovascular events, yet its genetic determinants remain poorly understood. Advances in deep learning have enabled the extraction of precise phenotypic measures from cardiac magnetic resonance (CMR) imaging, facilitating large-scale genome-wide association studies (GWAS) to uncover novel genetic insights into cardiovascular disease (CVD) mechanisms. Aim This study leverages deep learning and GWAS to identify genetic loci and biological pathways associated with aortic stiffness, using a novel pressure-independent measure derived from CMR imaging data in the UK Biobank. Methods We applied a pre-trained neural network to segment CMR images from 45,789 UK Biobank participants of European ancestry, quantifying ascending aorta diameter. A pressure-independent measure of aortic stiffness (β0) was calculated using an exponential pressure-diameter relationship, incorporating systolic and diastolic blood pressure measurements. GWAS was performed on β0, followed by stepwise conditional joint analysis to identify independent lead variants. Loci were defined as 500kb flanking regions centered on each variant. Putative effector genes were prioritized using a multi-step approach: variant-to-gene mapping, polygenic priority scoring, gene-based association testing, and nearest gene analysis. Pathway enrichment analysis was conducted to elucidate biological mechanisms underlying aortic stiffness. Results The GWAS identified 17 independent loci associated with aortic stiffness. Putative effector genes were resolved at 16 loci, including ELN and LTBP4, implicated in elastin fiber formation (p=1.9E-2); ELN, HAS2, and LTBP4, linked to extracellular matrix assembly (p=7.5E-3); and ULK4, ARHGAP24, ELN, HAS2, SVIL, ARHGAP22, CDH13, SMG6, and LTBP4, associated with regulation of cellular component organization (p=2.6E-2). Conclusion This study demonstrates the power of integrating deep learning and GWAS to uncover genetic determinants of aortic stiffness, highlighting roles for extracellular matrix and cellular organization pathways. These findings advance our understanding of aortic stiffness and provide a foundation for targeted mechanistic studies and precision medicine approaches in cardiovascular disease.Manhattan plot and gene prioritisation.