When Nicholas Kim was in eighth grade, his grandfather, the man who had taken him to basketball practice and taught him to tie his shoes, began to change. He developed Alzheimer’s disease, and the roles of caregiver and child quietly reversed. Kim, whose parents worked long hours as a doctor and a pharmacist, found himself at his grandfather’s home, bathing him, brushing his teeth, and even carrying him to the bathroom.
That experience led Kim to embrace science with a purpose. Now 22 and finishing his senior year at USC Viterbi’s Alfred E. Mann Department of Biomedical Engineering, he has channeled that grief into something remarkable: a landmark research paper that for the first time maps the genetics of how individual regions of the brain age —and why some of those regions are the very ones most ravaged by Alzheimer’s and dementia.
“A big part of why I’m studying biomedical engineering is I wanted to think of ways I could help him and others like him with the disease,” said Kim, whose grandfather died when he was 15.
He is advised by Andrei Irimia, associate professor of gerontology, quantitative & computational biology, biomedical engineering and neuroscience at USC.
Beyond a Single Number
Published in the journal GeroScience, the paper is titled “Deep Neural Networks and Genome-Wide Associations Reveal the Polygenic Architecture of Local Brain Aging.” Where previous studies assigned the brain a single aging score, Kim’s research asked a more precise question: which specific genes drive aging in which specific brain regions?
For about a decade, scientists have measured what they call “brain age,” an estimate of how old your brain appears on an MRI scan, which can differ from your actual age. A 40-year-old’s brain might, for instance, look 50, a gap can signal elevated risk for cognitive decline.
“We’ve treated brain age like a single number, almost like a GPA for your brain,” Kim said. “But just like a GPA, that single score hides a lot of nuances.”
What Kim and his team discovered is that the brain doesn’t age uniformly. Different regions age at different rates, and those differences aren’t random; They are driven by specific genes.
By analyzing MRI scans from 41,708 adults in the UK Biobank, a large British health database, the researchers divided the brain into 148 distinct regions and measured the aging rate of each one separately. They then scanned each participant’s DNA, testing more than 600,000 genetic variants, and identified which variants were linked to accelerated aging in which regions. The result: 1,212 significant genetic associations, a detailed genetic map of how and where the brain grows old.
Genes That Age the Brain and Genes That Protect It
The study identified both factors that speed up aging and those that protect against it. One gene in particular, KCNK2, which controls potassium channels that help regulate electrical signaling between neurons, was strongly associated with faster aging in brain regions that are especially vulnerable in Alzheimer’s disease.
On the other hand, variants in a gene called NUAK1, which helps maintain the structural skeleton of brain cells, were associated with a younger-appearing brain across wide areas of the cortex.
Kim is careful not to overstate what this means for any individual. “Carrying a risky genetic variant is like having a slightly heavier backpack,” he said. “It makes the climb harder, but it doesn’t decide whether you reach the top. Lifestyle, environment, vascular health, cognitive engagement, these all matter enormously.”
One of the study’s most significant findings is that the regions of the brain that age fastest correspond closely to the regions most devastated by Alzheimer’s disease and frontotemporal dementia.
Kim sees this as a meaningful confirmation rather than a surprise. “What we’re looking at are the genetic variants that are causing those places to age quicker,” he said. The finding suggests that accelerated local brain aging may be an early biological signal of the same vulnerability that eventually leads to neurodegenerative disease.
Could this research one day help a doctor identify who is at risk for dementia years before symptoms appear, or guide the development of targeted treatments? Possibly. “This is mostly a powerful research tool right now, not a diagnostic test,” he said. “There are a lot of barriers to moving to the clinical side. Maybe in decades.”
The Role of AI
None of this would have happened without artificial intelligence. Each MRI scan is a three-dimensional image made up of more than two million tiny cubes of data. No human could process that volume. Kim’s team designed and built a custom AI system, a 3D neural network, that learned to detect the subtle structural signatures of aging across every region of the brain simultaneously.
The entire project took about a year and a half and required a computer cluster of four servers running roughly 120 processors simultaneously.
“AI was essential because aging signals are very subtle,” Kim said. “We trained a neural network to learn the structural patterns associated with age, and that gave us the trait we needed to run the genetic study.”
An Undergraduate Among Ph.D.s
Kim found his way to Irimia’s lab through USC’s Center for Undergraduate Research in Viterbi Engineering (CURVE) program. Working alongside professors and doctoral students was daunting. “I felt out of my depth,” he said. “But that feeling of being in over your head is where you can learn the most. It’s a sign you’re swimming in the right ocean.”
Irimia said his mentee’s work impressed him greatly and points to something beyond Kim’s technical skills. “When early versions of his algorithms failed, he rebuilt them from first principles,” the professor said. “He teaches himself advanced mathematics when needed, and he persists until the problem is solved correctly.”
“It is highly unusual for an undergraduate to independently conceive, execute, and publish a multi-year, genome-wide neuroimaging study at this level,” Irimia said. “The scale, rigor, and originality of his work are comparable to that of a highly productive Ph.D. student.”
This is not Kim’s first time as a lead author. He previously published a first-author paper in the journal Neuroinformatics on the genetics of cortical thickness and brain structure. That work earned him the Barry Goldwater Scholarship, a prestigious national award for undergraduate researchers. He carries a 4.0 GPA and attends USC on a full-tuition Trustee Scholarship.
Kim is currently applying to medical school and hopes to pursue an MD-Ph.D., combining clinical practice with research.
Irimia predicts great things for him.
“Nicholas has the rare combination of computational depth, biological curiosity, and leadership ability,” Irimia said. “I expect him to become a physician-scientist who not only practices medicine but also shapes the future of how we understand and treat neurodegenerative disease.”
Kim’s co-authors include Ayati Mishra, a USC sophomore studying neuroscience; neuroscience Ph.D. students Owen M. Vega, and Nahian F. Chowdhury; Nikhil Chaudhari, a biomedical engineering doctoral student whose adviser is Irimia; lab manager Samuel D. Anderson; Kenneth H. Buetow, a professor of genetics at Arizona State University; Paul M. Thompson, director of the USC Imaging Genetics Center and professor of ophthalmology, neurology, psychiatry and the behavioral sciences, radiology, psychiatry, electrical and computer engineering, and biomedical engineering; and Irimia, who supervised the project.
Published on March 16th, 2026
Last updated on April 20th, 2026
Original source: am