Genome-wide analysis of subcortical aging identifies a spatially structured pattern of genetic associations.
TL;DR
This work presents the first genome-wide association study (GWAS) of subcortical LBA, estimated using a deep neural network applied to T1-weighted MRI scans from 41,957 cognitively normal participants in the UK Biobank, and identifies 14 significant single-nucleotide polymorphisms across nine independent loci.
Abstract
Local brain age (LBA) is a spatially resolved biomarker of brain aging that captures regional deviations from chronological age, yet its genetic architecture in the subcortex remains unexplored. Here, we present the first genome-wide association study (GWAS) of subcortical LBA, estimated using a deep neural network applied to T1-weighted MRI scans from 41,957 cognitively normal participants in the UK Biobank. We computed LBA across 14 subcortical structures and identified 14 significant single-nucleotide polymorphisms (SNPs) across nine independent loci. These variants map to genes involved in cellular homeostasis, gene regulation, and synaptic and developmental signaling. A prominent signal emerged at the 17q21.31 haplotype, encompassing MAPT-related regulatory architecture, with significant associations across all subcortical regions. Across loci, we observed a recurring spatial pattern in which effect sizes are relatively larger in metabolically central structures such as the pallidum and thalamus compared to limbic regions. Together, these findings support a spatially structured pattern of genetic associations in subcortical brain aging. This work supports subcortical LBA as a genetically informed phenotype and provides a framework for linking common genetic variation to region-specific vulnerability and resilience in neurodegenerative disease.