Hum Mol Genet. 2014 Oct 1;23(19):5294-302. doi: 10.1093/hmg/ddu228. Epub 2014 Jun 6.

Expression QTL-based analyses reveal candidate causal genes and loci across five tumor types.

Li Q¹, Stram A², Chen C³, Kar S⁴, Gayther S⁵, Pharoah P⁴, Haiman C², Stranger B⁶, Kraft P³, Freedman ML⁷.

Author information

1: Department of Medical Oncology, The Center for Functional Cancer Epigenetics, Dana Farber Cancer Institute, Boston, MA, USA Medical College of Xiamen University, Xiamen, China Program in Medical and Population Genetics, The Broad Institute, Cambridge, MA, USA.
2: University of Southern California, Los Angeles, CA, USA.
3: Department of Epidemiology, Harvard School of Public Health, Boston, MA, USA.
4: Strangeways Research Laboratory, University of Cambridge, Cambridge, UK.
5: Department of Preventive Medicine, Keck School of Medicine, University of Southern California, Log Angeles, CA, USA.
6: Institute for Genomics and Systems Biology, University of Chicago, Chicago, IL, USA.
7: Department of Medical Oncology, The Center for Functional Cancer Epigenetics, Dana Farber Cancer Institute, Boston, MA, USA Program in Medical and Population Genetics, The Broad Institute, Cambridge, MA, USA freedman@broadinstitute.org.

Abstract

The majority of trait-associated loci discovered through genome-wide association studies are located outside of known protein coding regions. Consequently, it is difficult to ascertain the mechanism underlying these variants and to pinpoint the causal alleles. Expression quantitative trait loci (eQTLs) provide an organizing principle to address both of these issues. eQTLs are genetic loci that correlate with RNA transcript levels. Large-scale data sets such as the Cancer Genome Atlas (TCGA) provide an ideal opportunity to systematically evaluate eQTLs as they have generated multiple data types on hundreds of samples. We evaluated the determinants of gene expression (germline variants and somatic copy number and methylation) and performed cis-eQTL analyses for mRNA expression and miRNA expression in five tumor types (breast, colon, kidney, lung and prostate). We next tested 149 known cancer risk loci for eQTL effects, and observed that 42 (28.2%) were significantly associated with at least one transcript. Lastly, we described a fine-mapping strategy for these 42 eQTL target-gene associations based on an integrated strategy that combines the eQTL level of significance and the regulatory potential as measured by DNaseI hypersensitivity. For each of the risk loci, our analyses suggested 1 to 81 candidate causal variants that may be prioritized for downstream functional analysis. In summary, our study provided a comprehensive landscape of the genetic determinants of gene expression in different tumor types and ranked the genes and loci for further functional assessment of known cancer risk loci.