Fig. 3
Functional annotations and signs of natural selection in reQTLs. a Total numbers of cis eQTLs, and proportions of reQTLs and constant eQTLs (ceQTL) that have similar regression slopes across all conditions. Results of the analysis of 134 individuals are shown unless indicated otherwise. Examples of a ceQTL and reQTL are shown in Supplementary Fig. 8a. b Forest plot of enrichment estimates of reQTL and ceQTL signals for each functional annotation with 95% confidence intervals (see also Supplementary Fig. 5b). Asterisks indicate annotations that improved the model likelihood in a stepwise procedure for the final best-fitting model. Bar plot shows the enrichment of the single most likely causal SNP per locus after fine mapping. The solid bars indicate significant enrichments after Bonferroni correction. c Signal of positive selection measured as the proportion of variants with high |iHS| (left panel), and median |SDS| (right panel), using the variant with the maximum value from each locus across all SNPs in high LD (r 2 > 0.8). Genome-wide null sets of variants matched to eQTL, ceQTL or reQTL were generated by resampling 10,000 sets of random SNPs that matched for MAF and LD (white bars). Error bars indicate minimum and maximum of the null distribution, and asterisks indicate the significant enrichment compared to the null (permutation test p < 10−4). d Illustration of reQTLs where the derived allele causes an increase (left panel) or decrease (right panel) in response amplitude compared to the ancestral allele. The increase or decrease of the response amplitude can be in both directions, e.g., reQTLs that amplify the induction or amplify the suppression of a gene are both considered as reQTLs with “increasing activity” of the derived allele and reQTLs that weaken the induction or suppression of a gene are both considered as reQTLs with “decreasing activity” of the derived allele. e Numbers of reQTLs with increased or decreased activity across all stimulated conditions, with a p-value from a binomial test
