Imaging genetics for complex disease in action
September 28, 2026
Nazlee Zebardast and colleagues trained an ML framework to learn glaucoma damage patterns from ~19,000 OCT scans of glaucoma patients, then used those learned endophenotypes to power multi-ancestry GWAS in ~39,000 UK Biobank participants.
The payoff — 43 genome-wide significant loci in cross-ancestry meta-analysis, over a third novel to glaucoma, and 11 putative causal genes pointing to Wnt-mediated outflow dysfunction and retinal ganglion cell vulnerability.
