Deep neural networks and genome-wide associations reveal the polygenic architecture of local brain aging.
Authors
Affiliations (8)
Affiliations (8)
- Alfred E. Mann Department of Biomedical Engineering, Viterbi School of Engineering, University of Southern California, Los Angeles, CA, USA.
- Ethel Percy Andrus Gerontology Center, Leonard Davis School of Gerontology, University of Southern California, Los Angeles, CA, USA.
- School of Life Sciences Center for Social Dynamics and Complexity, Arizona State University, Tempe, AZ, USA.
- Imaging Genetics Center, Mark and Mary Stevens Neuroimaging and Informatics Institute, University of Southern California, Marina del Rey, CA, USA.
- Alfred E. Mann Department of Biomedical Engineering, Viterbi School of Engineering, University of Southern California, Los Angeles, CA, USA. [email protected].
- Ethel Percy Andrus Gerontology Center, Leonard Davis School of Gerontology, University of Southern California, Los Angeles, CA, USA. [email protected].
- Department of Quantitative & Computational Biology, Dornsife College of Arts and Sciences, University of Southern California, Los Angeles, CA, USA. [email protected].
- Centre for Healthy Brain Aging, Institute of Psychiatry, Psychology & Neuroscience, Department of Psychological Medicine, King's College, London, England, UK. [email protected].
Abstract
Local brain age (LBA) is a regional metric of brain aging that offers a spatially resolved alternative to global brain age, but whose genetic basis is unexplored. This study reports the first genome-wide association study of cortical LBA, as estimated by a deep neural network from the T<sub>1</sub>-weighted magnetic resonance images of 41,708 cognitively normal adults in the UK Biobank. We identified 1,212 single-nucleotide polymorphisms (SNPs) significantly associated with LBA in at least one brain region. Genes mapped to these SNPs are involved in developmental, metabolic, immune, and cytoskeletal pathways. Dimensionality reduction of SNP association profiles uncovered three clusters linked to morphogenetic, cytoskeletal, and immuno-epigenetic processes, helping to relate neuroanatomic, immunosenescent and epigenetic mechanisms of brain aging. Top variants are mapped to KCNK2, NUAK1, GMNC, MSL2, and to other genes implicated in neurodevelopment and neurodegeneration. Spatial clustering of LBA-associated variants in default mode, limbic, and motor network regions parallel regional vulnerability to Alzheimer's disease and frontotemporal dementia. These findings establish a polygenic architecture for regional brain aging and support LBA as a genetically informed phenotype for studying aging-related neurodegeneration. Our results suggest that cortical aging is not governed by isolated loci but by coordinated genetic programs-rooted in development, metabolism, and cellular structure-that confer lifelong patterns of regional brain vulnerability and resilience. This first genetic dissection of spatially specific brain aging reveals a polygenic landscape of coordinated genetic programs, developmentally encoded and metabolically maintained during senescence. This study reframes aging as an anatomically specific genetic process reflecting the varying structural vulnerabilities observed across neurodegenerative diseases.