• Less distinct boundaries between active and inactive genome regions, a pattern the authors call "increased compartment mingling"
• Fewer short-range DNA interactions paired with more long-range contacts
• Weaker links between genes and their nearby regulatory switches

Cells with greater mingling showed lower overall gene activity. The structural changes aligned with reduced programs for neurons and synapses, alongside shifts in metabolism and stress responses. Microglia — the brain's resident immune cells that maintain brain health and respond to damage — showed links to senescence programs.

How does the AI model fit in?

Hicformer takes in DNA sequence information together with 3D genome folding patterns and predicts gene activity across different cell types. Xinyue Lu, a doctoral student at CMU who co-led the work, called the model a "computational test bed" for exploring how folding changes might alter gene expression.

Yang Zhang, a project scientist at CMU and co-lead author, described the broader benefit: "Measuring gene activity and genome folding in the same cell allows us to directly connect chromosome structure with disease-related gene programs."

The model's predictions helped researchers flag specific regulatory DNA regions for future experimental tests.

What does this add to Alzheimer's biology?

Research on Alzheimer's has long centered on amyloid-beta plaques and tau tangles, the protein clumps that define the disease. The new findings argue for adding chromatin — the DNA-and-protein packaging that organizes the genome — to the picture.

"We know the classic hallmarks of Alzheimer's disease — accumulation of amyloid-beta plaques and tau tangles — but our results establish higher-order chromatin alterations as a component of the molecular pathology associated with the disease, which currently affects seven million Americans, a number that continues to grow," said Mathys, who directed the Pitt arm of the research.

When the team overlaid the molecular changes onto intact brain tissue, the genome reorganization also tracked with differences in how cells distributed themselves.

What's next?

The researchers frame their findings as a foundation rather than a treatment. Future studies can now investigate whether any of the flagged regulatory regions help drive Alzheimer's progression, and whether those regions could eventually become targets for new therapies.

Additional collaborators came from the Broad Institute of MIT and Harvard, UCLA, the University of Washington, and the Rush Alzheimer's Disease Center. Funding came from the National Institutes of Health.