Testing the utility of an integrated analysis of copy number and transcriptomics datasets for inferring gene regulatory relationships.

Abstract

Correlation patterns between matched copy number variation and gene expression data in cancer samples enable the inference of causal gene regulatory relationships by exploiting the natural randomization of such systems. The aim of this study was to test and verify experimentally the accuracy of a causal inference approach based on genomic randomization using esophageal cancer samples. Two candidates with strong regulatory effects emerging from our analysis are components of growth factor receptors, and implicated in cancer development, namely ERBB2 and FGFR2. We tested experimentally two ERBB2 and three FGFR2 regulated interactions predicted by the statistical analysis, all of which were confirmed. We also applied the method in a meta-analysis of 10 cancer datasets and tested 15 of the predicted regulatory interactions experimentally. Three additional predicted ERBB2 regulated interactions were confirmed, as well as interactions regulated by ARPC1A and FANCG. Overall, two thirds of experimentally tested predictions were confirmed.