Skip to main content
Fig. 5 | Genome Biology

Fig. 5

From: Cancer expression quantitative trait loci (eQTLs) can be determined from heterogeneous tumor gene expression data by modeling variation in tumor purity

Fig. 5

Conventional model eQTLs for breast cancer risk GWAS significant genetic variants, which cannot be attributed to cancer cells by the interaction model, are all statistically significant with concordant directionality in GTEx LCLs, breast, or fibroblasts. a Scatterplot of eQTLs identified for breast cancer risk variants from a meta-analysis of GWAS data [6]. -log10 P values for significant eQTLs (FDR < 0.05) using a conventional model in TCGA breast cancer tumors are shown on the x-axis. -log10 P values from the interaction model for these same eQTLs are shown on the y-axis. Significance thresholds (P < 0.05) are shown for the interaction model (green and purple dashed lines). 15 of these 24 eQTLs were no longer significant (FDR > 0.05) when the possibility of these eQTLs arising from tumor-associated normal cells was modeled. b P values and effect sizes for the 9 eQTLs in (a) that were no longer even nominally significant (P > 0.05) when the interaction model was used. For each eQTL, the effect size is represented by the red-blue divergent color scale and the P value by the size of the point; these are shown for TCGA bulk breast tumor (i.e., the conventional model) and transformed fibroblast, breast tissue, and LCLs from GTEx. All of these eQTLs are evident in at least one of these GTEx tissues, with 100% concordant direction of effect as identified from TCGA breast bulk tumor expression using the conventional model. Points are shown for associations with P < 0.05

Back to article page