A SHARED GENETIC SIGNAL ACROSS BRAIN AGE PREDICTION MODELS REVEALS ROBUST MARKERS OF BRAIN HEALTH

Deviations from normative brain ageing trajectories are linked to a wide range of adverse health outcomes. Multiple brain age prediction models have been developed to quantify these deviations, using diverse neuroimaging modalities, machine learning approaches, and age ranges. However, it remains unclear whether these models converge on a shared genetic liability, and whether this shared signal could serve as a more sensitive marker of brain health than any single model alone. We first conducted a new brain age gap (BAG) GWAS in 60,735 individuals from 29 cohorts worldwide and then applied genomic structural equation modelling to examine the shared genetic variance across this GWAS and five previously published BAG GWASs using a single latent BAG factor (30 cohorts total). All six BAG GWASs loaded onto a single factor, explaining 63% of the total genetic variance. Nineteen independent SNPs associated with the BAG factor, including four novel associations. The BAG factor was genetically correlated with multiple traits, among which blood pressure, smoking, longevity, autism, and sleep showed putatively causal effects. Polygenic scores (PGS) for the BAG factor were associated with BAG phenotypes already in childhood, with stronger associations observed in adulthood. Phenome-wide analyses showed that the BAG factor PGS captured associations with more health traits than individual BAG PGSs. Overall, these findings suggest that the shared genetic signal may provide a more robust marker of brain health than any single model alone.