A new AI model reveals that Alzheimer’s disease risk is written differently in male and female DNA.
Why do women make up nearly two-thirds of all Alzheimer’s patients? For years, clinical consensus blamed the simple fact that women live longer. This new study suggests a more uncomfortable truth: the genetic architecture of the disease itself is fundamentally different depending on biological sex.
By ignoring these differences, drug developers may be designing therapies that work for one sex while failing the other.
Researchers built a Transformer model called STAGE-AD to analyze genetic data from 449,335 participants of European ancestry. The model achieved a strong primary precision-recall area under the curve of 0.895. However, when researchers held out the well-known APOE gene region, performance slid to 0.681. This drop reveals a crucial limitation: outside of the usual genetic suspects, predicting individual risk remains a steep challenge.
The genetic divide
The analysis uncovered distinct genetic signatures that traditional, pooled models miss. This challenges the industry habit of treating male and female genomes as functionally identical in dementia research.
- The model identified 8 female and 3 male novel genome-wide-significant signals.
- Independent validation in the HUNT and MVP cohorts confirmed 236 genes across 106 loci.
- Estimated heritability on the liability scale was 11.42% for females, compared to 9.23% for males.
These numbers confirm that genetics play a larger, more complex role in female Alzheimer’s risk than in male risk. This gap should force researchers to rethink how they design clinical trials, as pooling sexes may dilute therapeutic signals and sink promising drugs.
Different cells drive risk
The real analytical breakthrough lies in the cellular pathways. Ablation tests showed that female-biased risk predictions depended heavily on glial cells, which manage the brain’s immune response. Conversely, male-biased predictions relied on endothelial cells and oligodendrocytes, which control blood vessels and nerve insulation.
To confirm these findings, the team mapped these genetic chains against brain tissue from 579 donors. The biology suggests that female Alzheimer’s is primarily an inflammatory and support-cell issue, while male Alzheimer’s is a vascular and structural one.
This means a drug targeting neuroinflammation might succeed in women but fail in men. We must stop treating Alzheimer’s as a single, uniform disease.
However, we must view these findings with caution. The study relied entirely on cohorts of European ancestry, meaning these sex-specific pathways might not translate to other populations. Until we validate these models globally, we have only mapped one piece of the puzzle.
Read the full study in medRxiv.
