Fig. 3: A joint predictive model of gene expression identifies tissue-specific contributions of age and genetics to transcript levels.
From: Tissue-specific impacts of aging and genetics on gene expression patterns in humans
a A schematic of our multi-SNP gene expression association model incorporating sample age. Common SNPs around each gene g are used in combination with an individual's age to predict expression within tissue t. Using this trained model, variation in gene expression can be separated into three parts: the components explained by genetics (\({R}_{{{{{{{{\rm{genetics}}}}}}}}}^{2}\) or h2), by age (\({R}_{{{{{{{{\rm{age}}}}}}}}}^{2}\)) and by all other factors (\({R}_{{{{{{{{\rm{environment}}}}}}}}}^{2}\)). b Proportion of each gene's expression variance explained by age and genetics. c Plot of normalized expression vs age for four genes with age-correlated expression. Line shows fitted βage from regularized linear model. d Point estimates of the mean \({R}_{{{{{{{{\rm{age}}}}}}}}}^{2}\) and h2 for each tissue, error bar indicating standard error for the estimate. e The tissue specificity score of R2 across 27 tissues for each gene from either age or genetics. Center line of the boxplot indicates median, box limit indicates first and third quartiles, points on both ends indicate minima and maxima. P-value is obtained from two-sided paired samples t-test with n = 812 genes.
