By combining artificial intelligence-based imaging analysis with large-scale human genetics, the team from UC San Francisco and the Broad Institute of MIT and Harvard has uncovered early genetic signals that shape aortic valve function long before clinical disease develops.
The findings, published in Nature Genetics, point toward a future in which aortic stenosis could be detected, and potentially intercepted, years earlier.
Using deep learning models trained on cardiac MRI, the team extracted three continuous measures of aortic valve function—peak velocity, mean gradient, and aortic valve area—from nearly 60,000 UK Biobank participants who did not have diagnosed valve disease.
These AI-derived measurements capture subtle differences in valve performance that are invisible in routine clinical care.
A combined multi-trait analysis identified 166 genetic loci associated with aortic valve function or aortic stenosis, demonstrating that the boundary between “normal” valve variation and disease is genetically continuous.
The study showed substantial genetic correlation between valve function in healthy individuals and clinical aortic stenosis.
Gradient-based valve measures had a correlation of 0.64 with disease risk, while aortic valve area showed a correlation of 0.50.
In practical terms, this means that many of the same genetic variants that slightly alter valve function in healthy people also increase the likelihood of developing clinically significant stenosis later in life.
【MORE】