Blood immune cells changed with age at different rates in women and men among 1 828 healthy people
Blood immune cells changed with age at different rates in women and men among 1 828 healthy people
In an article published on August 21, 2026, researchers combined single-cell data from 1 828 healthy people aged 19 to 97 years. Among 3,8 million immune cells from blood samples, they identified two periods of life when the activity of a particularly large number of genes changed. In external datasets, separate models for women and men predicted chronological age from these profiles more accurately than a single combined model.
The immune system changes with age, but averaging the trajectory across all people obscures differences in when these changes occur. The authors combined four public datasets and generated an aggregate gene activity profile for every donor within each cell type. Reducing the data to this type of profile gave each donor equal weight, regardless of the number of cells in the sample.
For each age, the researchers then compared two adjacent twenty-year groups. The analysis identified periods of life when a particularly large number of genes differed between the groups. The first peak occurred at about 40 years of age and was driven mainly by CD4 T-cells, which coordinate the immune response. After 60 years of age, the most pronounced changes occurred in CD8 T-cells, which destroy infected and altered cells. The same two peaks remained when the researchers changed the width of the age windows, adjusted the statistical thresholds, and randomly reduced the dataset.
Separating the data by sex changed the pattern. In women, signs of CD8 T-cell activation increased throughout life, while later changes also affected CD4 T-cells, NK cells, and B-cells. In men, the earlier fluctuations were concentrated in CD4 T-cells and cellular energy metabolism pathways. Women and men of the same chronological age may therefore be at different stages of these immune programs.
The team developed neural network “clocks,” which are models that predict chronological age from gene activity. In the model for women, the group of features that increased continuously with age contributed most to the prediction. In the model for men, the group that increased earlier contributed most. The separate models retained these temporal patterns and performed better on new datasets.