Fig. 3

Age-specific genetic regulation across the genome. a Number of genes with at least one significant eQTL (eGenes) for uncorrected analysis (0 hidden factors) and analysis corrected for one up to 15 hidden factors. We detected 1326 and 1264 eGenes at age 70 and 80 (FDR ≤ 1%), respectively. The depletion of eGenes at age 80, relative to age 70, is statistically significant (exact McNemar’s test; OR = 0.81, P=5.8×10⁻³). Dashed line indicates the number of hidden factors that maximizes discovery at each age. b Proportion of eGenes discovered (FDR ≤ 1%) at age 70 (80) that validated (FDR ≤ 10%) at age 80 (70). The validation proportion of eGenes discovered at age 70 is significantly smaller than the proportion at age 80 (binomial proportion test; \(\pi ^{val}_{70}-\pi ^{val}_{80} = -2.7\%, P = 3.3 \times 10^{-3}\)). c Difference in expression cis-heritability between ages. The decrease in average cis-heritability with age is statistically significant (Wilcoxon signed rank test; \(median(h^{2}_{80}-h^{2}_{70})= -1.4\%\), P=5.56×10⁻¹⁴). d Scatter plot of eQTL effect sizes at each age for genes that are eGenes in at least one age group. We observed a strong correlation of the fixed effect sizes between the two ages (Spearman’s ρ_β=0.97). Blue line represents linear regression fit