Skip to main content
Fig. 3 | Genome Biology

Fig. 3

From: Tejaas: reverse regression increases power for detecting trans-eQTLs

Fig. 3

Tejaas identifies many thousands of putative trans-eQTLs in GTEx data. In each of the 49 GTEx tissues, we applied the KNN confounder correction and calculated genome-wide reverse regression p values with Tejaas. Cis genes within ± 1Mb of the candidate SNP were excluded from the regression. From the genome-wide significant SNPs (p<5 × 10−8), we selected the strongest in each LD region as lead trans-eQTLs, removing other SNPs in strong LD (r2≥0.5) with the lead SNP. a Number of lead trans-eQTLs discovered per tissue, on a logarithmic scale. For GTEx tissue abbreviations, see Additional file 1: Appendix 2. The dotted line indicates the cutoff used for choosing tissues for enrichment analysis. b Proportion of trans-eQTLs discovered in a given number of tissues (excluding brain tissues). Seventy percent of the lead trans-eQTLs are not in strong LD with any lead trans-eQTL from another tissue. c Number of lead trans-eQTLs discovered in a tissue (log scale) versus the number of samples for that tissue (tissue colors as in a). d About a fifth of the trans-eQTLs have detectable cis-effects. Number of lead trans-eQTLs versus the number of discovered lead trans-eQTLs that also happen to be cis-eQTLs in GTEx consortium analysis [6]. Tissue colors as in a, radii scale with sample sizes (legend). (see Fig. 4a for corresponding enrichments.) e Representative examples of quantile-quantile plots for artery aorta (ARTAORT) and EBV-transformed lymphocytes (LCL) with their negative controls (dashed), obtained by randomly permuting the sample IDs of genotypes. f Representative examples trans-eQTL maps for ARTAORT and LCL, with genomic positions of trans-eQTLs (x-axis) against the genomic positions of their target genes (y-axis). The diagonal band (blue) corresponds to cis-eQTLs

Back to article page