Fig. 5: Estimating the age of individual CD8⁺ T cells based on their transcriptomic signatures.

a Cross-validation age predictions for CD8⁺ T cells from each participant based on scRNA-seq data using a fitted mixed-effect elastic net (MEEN) model. Distribution of predicted cell ages for each donor is shown by violin plot (blue, cross-sectional donors; red, longitudinal donors). Donors’ chronological age is on the x-axis. Trend lines were from the lineage regression model for which the t distribution with n-2 degrees of freedom was used to evaluate if a slope was significant for a particular variable (2 sided t-test). The data are in Supplementary Data 5a. b Left, cross validation of predictions for donors older than 70 years (Old) and younger than 30 years (Young). Right, test predictions for an independent dataset from healthy donors older than 70 years (Old) and younger than 30 years (Young) using public scRNA-seq datasets for CD8⁺ T cells. The box-whisker plots in a and b: the center line is the median; the box is from the 25^th to the 75^th percentile. The upper or lower whisker extends from the hinge to the 1.5 x IQR (distance between the first and third quartiles) from the hinge for up and low, respectively. P values were calculated by 2-sided T-test. The data are in Supplementary Data 5b. c Overlap of age-associated genes in CD8⁺ T cells between MELR and MEEN models. The percentages of overlaps were incremented by the top 100, 200, and 300 genes identified by MELR based on FDR and coefficient of age. Percentages are grouped by if the genes were up or downregulated across aging as identified by the MELR method. The data are in Supplementary Data 5c. d Scatter plot of the coefficient of age (CO) of age-associated genes identified by MELR and the percentages of cells expressing each gene. The average change (CO) and percentage of each six groups are presented.