The researchers asked whether Alzheimer’s disease is associated not only with changes in gene activity but also with cell-type-specific changes in the three-dimensional folding of the genome that may help explain those expression patterns.
The research question and why it matters
The researchers asked whether Alzheimer’s disease is associated not only with changes in gene activity but also with cell-type-specific changes in the three-dimensional folding of the genome that may help explain those expression patterns.
Earlier single-cell studies documented cell-type-specific gene-expression and chromatin-accessibility changes in Alzheimer’s disease, while other work developed methods for measuring genome folding and RNA in the same cell. This study connected those layers in the same nuclei and placed them in spatial tissue context, extending rather than replacing established amyloid, tau, inflammatory and genetic evidence.
What researchers found
Across multiple brain-cell types, Alzheimer’s samples showed a higher ratio of long- to short-range chromatin interactions and more contact between normally separated active and inactive genome compartments. Gene-to-regulatory-element contacts were also reorganized. These structural patterns coincided with cell-type-specific changes in gene programs, including elevated senescence-related expression in a subset of microglia. The modeling analysis found that adding three-dimensional genome features improved predictions of disease-associated gene-expression differences.
Key results from the tested systems
brain donors
Ten Alzheimer’s and ten age-matched non-Alzheimer’s donors were compared.
profiled nuclei
Gene expression and three-dimensional chromatin contacts were jointly analyzed.
two molecular layers
GAGE-seq measured RNA and genome architecture together.
study window
The design cannot determine whether the folding changes came before disease.
How the research worked
The team used GAGE-seq to measure RNA expression and chromatin contacts in the same nuclei from prefrontal-cortex tissue. They classified neuronal and non-neuronal cell types, compared Alzheimer’s and non-Alzheimer’s donors, integrated matched chromatin-accessibility and spatial-transcriptomic information, and built Hicformer, a deep-learning system that tested how much DNA sequence and genome folding contributed to predictions of gene activity.
How to interpret this design
The result is conditional on the model structure, inputs, boundary conditions and scenarios chosen by the researchers. Agreement with known observations strengthens confidence, but a projection is not a direct observation of the future or the inaccessible past.
Because the research involved people, it speaks directly to the participants and outcomes measured. It may still apply differently to people outside the eligibility criteria, age range, clinical setting, geography or follow-up period.
What strengthens or limits the finding?
The study measured gene activity and three-dimensional chromatin contacts in the same nuclei, included matched pathology information, integrated independent molecular and spatial data, and compared results with prior work. Its 20-donor postmortem design is observational and cannot determine temporal order or causation.
The result is meaningfully informative, but identifiable limitations could alter the size, reach or causal interpretation of the finding.
Funding and disclosure context
The recorded funding source is: Grants from the U.S. National Institutes of Health; individual award numbers were not listed in the institutional report reviewed for this page. The recorded conflict information is: Zhijun Duan is an inventor on a University of Washington provisional patent application covering the GAGE-seq protocol; all other authors declared no competing interests. Funding or a disclosed relationship does not by itself invalidate a result, but it is relevant when judging design choices, analysis and the need for independent replication.
What it means
The results add genome folding to the molecular layers that researchers can investigate in Alzheimer’s disease. They point to specific chromatin interactions and cell states for experimental follow-up, but they do not yet provide a diagnostic test, a treatment target or evidence that changing genome structure would improve memory or slow disease.
Deeper analysis
Genome folding can regulate access
DNA is packed into the nucleus in loops and compartments. Regions brought close together can influence gene activity even when they are far apart along the chromosome, so a change in nuclear organization is biologically meaningful without changing the underlying DNA sequence.
The donor is the true independent unit
Tens of thousands of nuclei give detailed cell-type resolution, but they came from only 20 people. Cell counts improve measurement precision within tissue; they do not erase uncertainty about how broadly the donor-level pattern generalizes.
Association and mechanism are different stages
The study connects chromatin architecture with altered gene programs, but postmortem comparisons cannot establish direction. Experiments that perturb specific contacts in living cell or animal systems are needed before calling them disease mechanisms.
AI was an analytical component
Hicformer tested whether genome sequence and folding information help predict expression. Its success supports a relationship between these data layers; it is not evidence that an AI system can diagnose Alzheimer’s or identify a treatment on its own.
What it does NOT prove
- It does not show that altered genome folding begins before Alzheimer’s disease or causes its symptoms.
- It does not establish that the identified chromatin patterns are unique to Alzheimer’s rather than aging, inflammation or end-stage brain changes.
- It does not validate Hicformer as a clinical diagnostic or treatment-selection system.
- It does not demonstrate that reversing any chromatin contact will restore brain-cell function.
- It does not represent living-brain measurements or track changes within the same person over time.
Important limitations
- Only 20 donors were included, so donor-level biological variation can matter even though thousands of nuclei were measured.
- Postmortem tissue provides a late snapshot; disease progression, treatment, cause of death and postmortem interval can influence molecular measurements.
- Many single cells from one donor are not equivalent to the same number of independent people.
- The analysis focused on prefrontal cortex and may not describe other brain regions or early disease stages.
- Deep-learning predictions depend on the training data and do not by themselves identify a causal biological mechanism.
- The GAGE-seq method is covered by a provisional patent application involving one author, a disclosed interest relevant to the platform.
How this fits with previous research
Earlier single-cell studies documented cell-type-specific gene-expression and chromatin-accessibility changes in Alzheimer’s disease, while other work developed methods for measuring genome folding and RNA in the same cell. This study connected those layers in the same nuclei and placed them in spatial tissue context, extending rather than replacing established amyloid, tau, inflammatory and genetic evidence.
Questions still unanswered
- Which three-dimensional genome changes occur early enough to contribute to disease rather than follow it?
- Do the same patterns appear in larger, more diverse donor cohorts and other affected brain regions?
- Which altered regulatory contacts change cell behavior when manipulated in living models?
- How much of the signal reflects Alzheimer’s pathology versus normal aging or other dementias?
- Can accessible biomarkers ever capture these brain-tissue changes without a biopsy?
Relevant U.S. government resources
These resources serve different purposes. A registry can verify what researchers planned, a repository can locate government-funded work, and an agency page can supply authoritative background. None automatically proves that this paper's conclusion is correct.
PubMed record search ↗
Federal biomedical-index search keyed to this paper's DOI or title. It can confirm indexing and expose linked identifiers when a record is available; the journal paper remains the primary source.
Understanding clinical research ↗
NIH background on how clinical research is designed, reviewed and interpreted. This is contextual guidance, not independent confirmation of the study's result.
Reuse note: Facts and discoveries are summarized here in original language. We link to government material instead of copying it wholesale, and we do not reuse agency logos, photographs, charts or third-party material unless the specific reuse rights are verified.
Alzheimer’s brain cells showed changes in how their DNA was folded
This review was developed from the source record below and, when separately available, the primary paper or government report. The summary and analysis on this page are original editorial writing.
- Source organization
- Carnegie Mellon University
- Source type
- University
- Authors
- Yang Zhang, Xinyue Lu, Alexander K. Kunisky, Shahul Alam, Junjie Tang, Ruochi Zhang, Shike Wang, Han Zhang, Jude Baroudi, Walid Ichcho, Deyong Jia, Sahar Ghorbanikalateh, Sahel Ghorbanikalateh, Shihan Wang, David A. Bennett, Hansruedi Mathys, Zhijun Duan and Jian Ma
- Journal / report
- Science
- Publication date
- July 23, 2026
- DOI
- 10.1126/science.adz1652
- PMID
- 42490473
- Institution
- Carnegie Mellon University, University of Pittsburgh, Broad Institute of MIT and Harvard, UCLA, University of Washington, Rush Alzheimer’s Disease Center and Fred Hutchinson Cancer Center
- Funding
- Grants from the U.S. National Institutes of Health; individual award numbers were not listed in the institutional report reviewed for this page
- Conflicts
- Zhijun Duan is an inventor on a University of Washington provisional patent application covering the GAGE-seq protocol; all other authors declared no competing interests
- Open access
- Yes
- Reuse approach
- Study design and results summarized independently from Carnegie Mellon University, the peer-reviewed paper and the National Library of Medicine record; no source wording, figures, tables, photographs or illustrations reproduced.
AI-assisted editorial process: AI tools helped organize sources and draft this review. The linked research records—not AI output—are the evidence. Publication standards and corrections are publisher-directed. Read our AI transparency policy.