Preprint: age data from 490 donors used to map gene relationships and test predictions in skin cells
Preprint: age data from 490 donors used to map gene relationships and test predictions in skin cells
A preprint posted on bioRxiv on 23 October 2026 reports on cultured fibroblasts, connective tissue cells from the skin, from 490 donors aged 20 to 70. The authors ordered gene expression profiles by donor age, used this sequence to map directional relationships between genes, and tested some of the predictions by increasing the expression of individual genes in cells from 55-year-old donors.
Gene expression was measured once for each donor. Comparing people of different ages identifies genes whose expression changes with age, but this alone leaves too many possible relationships to test. The authors averaged the profiles of donors at each age in whole years, producing a sequence of 51 age points.
The CCM method compares these age trajectories: if the trajectory of one gene can reconstruct that of another, the method proposes a directional relationship to test in cells. After filtering, the map retained the top 5% of relationships, approximately 1,8 million. The authors selected genes from different parts of this list for a laboratory library and control experiments.
After quality control, the single-cell RNA analysis included 266 perturbations. In each experiment, the team increased the expression of a selected gene, causing cells to transcribe it more actively, and measured whether expression changed in the gene identified by the map. The measured response agreed with the CCM prediction for 53,8% of the relationships tested.
The genes were delivered into cells using a virus. A control using GFP, a fluorescent marker delivered by the same virus, showed that delivery itself altered the expression of many genes: expression changed for 10 536 genes in cells receiving GFP. After excluding this delivery-related effect, agreement was 30%; two comparison methods tested under the same conditions yielded 22,0% and 22,3%.
The team also measured how the perturbations shifted cell state. Algorithms were trained to distinguish control cells from donors aged 25 and 55 using RNA measurements and Cell Painting, which stains cellular structures to measure cell shape and internal organization. After the expression of some genes was increased, the algorithms more often classified cells from 55-year-old donors as belonging to the 25-year-old group. Scores for the same perturbations from the two measurement approaches had a correlation coefficient of 0,52.
The map identifies gene pairs for experiments, and the perturbation results help determine which hypotheses warrant further testing.