Fig. 5: Causal relationship between eQTL variants, TEs and genes.

a Barplot representing the mean probability for each of the three models in normal and tumor. We observe significantly more causal cases in tumor compared to normal (two-sided Wilcox P-value < 2e−16). b Barplot representing the model substitutions for the 9528 tumor triplets from normal to tumor. Independent models tend to shift to a causal in tumor. This is true also for the reactive models in normal but to a much smaller extent. c Barplot representing the number of triplets that do not switch models, that switch to a causal model or that switch to reactive/independent from normal to tumor. The majority of triplets do not switch models between normal and tumor. However, 2584 triplets are switching to a causal model making the corresponding TEs potential drivers of gene expression. d Each point represents a TE-gene for each of the 2584 tumor triplets. All points are significant in tumor but not in normal (gray points). We observe that in most cases, TEs are positively correlated with genes except for a few cases. Most cancer driver genes have no significant correlation with any TE in normal indicating that for most part, TEs impact them specifically in tumor. Source data are provided as a Source Data file.