Population genomics of four-dimensional cardiac motion reveals non-myocyte regulatory programmes
File(s) supplement.pdf (8.97 MB) nature_comms.pdf (34.48 MB)
Supporting information
Accepted version
Author(s)
Type
Journal Article
Abstract
Genome-wide studies of cardiac structure and function have relied on global imaging metrics that fail to
capture the full complexity of myocardial behaviour, leaving the genetic and environmental architecture of motion largely unexplored. Here we apply unsupervised deep learning to four-dimensional cardiac imaging from 78,113 adult participants in the UK Biobank to derive latent representations of left-ventricular motion. Genome-wide association analysis of these traits reveals 39 genetic loci, two-thirds undetected by conventional indices, and exposes distinct cell-type-specific regulatory programmes validated across single-cell atlases of the humanheart. Conventional traits resolve to cardiomyocyte and conduction-system regulatory elements, while motion-specific variants converge on non-myocyte programmes encompassing autonomic neuronal, endocardial, fibroblast, and endothelial compartments. All of Us, FinnGen and GPMap data provided locus level corroboration. Motion models also allowed mapping of how rare variants and environmental exposures associate with complex patterns of spatio-temporal function in the left ventricle.
capture the full complexity of myocardial behaviour, leaving the genetic and environmental architecture of motion largely unexplored. Here we apply unsupervised deep learning to four-dimensional cardiac imaging from 78,113 adult participants in the UK Biobank to derive latent representations of left-ventricular motion. Genome-wide association analysis of these traits reveals 39 genetic loci, two-thirds undetected by conventional indices, and exposes distinct cell-type-specific regulatory programmes validated across single-cell atlases of the humanheart. Conventional traits resolve to cardiomyocyte and conduction-system regulatory elements, while motion-specific variants converge on non-myocyte programmes encompassing autonomic neuronal, endocardial, fibroblast, and endothelial compartments. All of Us, FinnGen and GPMap data provided locus level corroboration. Motion models also allowed mapping of how rare variants and environmental exposures associate with complex patterns of spatio-temporal function in the left ventricle.
Date Acceptance
2026-10-23
Citation
Nature Communications
ISSN
2041-1723
Publisher
Nature Portfolio
Journal / Book Title
Nature Communications
Copyright Statement
Copyright This paper is embargoed until publication. Once published the Version of Record (VoR) will be available on immediate open access.
License URL
Publication Status
Accepted
