Shared genetic architecture of brain age gap across 30 cohorts worldwide.

Deviations from normative brain ageing trajectories are linked to a wide range of adverse health outcomes. A number of brain age prediction models have been developed, based on various neuroimaging modalities, machine learning algorithms, training samples, and age ranges. However, it remains unknown whether these models converge on a shared genetic liability, and whether capturing this shared signal could provide a more sensitive marker of brain health than any single model alone. We first conducted a new brain age gap (BAG) GWAS in a sample of 60,735 individuals across 29 cohorts worldwide, and then applied genomic structural equation modelling to examine the shared genetic variance between five prior BAG GWASs and our new analysis, using a single latent BAG factor (30 cohorts overall). All six BAG GWASs loaded onto a single factor, explaining 63% of the total genetic variance. We identified 19 independent SNPs associated with the BAG factor, including four novel associations. Genetically, the BAG factor was positively correlated with multiple traits, with blood pressure, smoking, longevity, autism, and sleep showing putatively causal effects. A polygenic score (PGS) for the BAG factor showed associations with phenotypic BAGs already in childhood, with stronger links observed in adulthood. Phenome-wide association analyses indicated that BAG factor PGS captured associations with more health traits than individual BAG PGSs. Our findings underscore the importance of considering the shared variance across different BAG constructs to identify robust correlates of poor brain health.