The flexibility of the linear mixed model framework allows variancePartition to analyze cross-individual variation within each tissue. We note again that since the variance is analyzed within multiple subsets of the data, the total variation explained no longer sums to 1 here. While variation across individuals explains only a median of 2.3% of variation when all tissues types are considered together, there is substantial variation across individuals within each tissue separately (Fig. 5 b). Cross-individual variation is highest in blood (median 60.3%), while skin (36.5%), blood vessel (22.5%), and adipose tissue (17.7%) exhibit lower cross-individual variation. The fraction of variation explained by individual within each tissue is directly related to the probability of each gene having a cis-eQTL within the corresponding tissue (Fig. 5 c). This association is not as strong as in other datasets likely due to the smaller number of individuals and to the relatively small fraction of variation across individuals in adipose tissue.
