Mapping eQTL-target gene associations in tumor tissue presents additional analytical challenges. Tumors acquire frequent genetic and epigenetic alterations, which can substantially affect gene expression. For example, somatic copy number changes and DNA methylation status are known to strongly influence transcript abundance in tumors (Curtis et al., 2012; Portela and Esteller, 2010). Consequently, the effect of these somatic alterations may obscure the association between germline genetic polymorphisms and gene expression. The creation of publicly available large-scale datasets, such as the Cancer Genome Atlas (TCGA), and Encyclopedia of DNA Elements (ENCODE) provide comprehensive catalogs of multiple data types performed on the same set of samples. In this study, we use these resources to develop a general method that models transcript levels as having inputs from germline and somatic factors.
