Tapping microRNA regulation networks through integrated analysis of microRNA-mRNA high-throughput profiles

Understanding the biological relevance and context of microRNA (miRNA) regulation of target mRNAs is difficult to ascertain because an individual miRNA aids simultaneously in the regulation of hundreds of mRNAs in a cell. With the increasing availability of large public datasets that profile both mRNA and miRNA expression levels from the same samples, it is possible to apply robust statistical methods to identify global negative correlations in miRNA and target mRNA expression. Using a dataset from The Cancer Genome Atlas as a case study, we show how to use linear regression analysis followed by permutation-based false discovery rate to assign high statistical power to pair-wise negative correlations of miRNA and mRNA expression. Used in conjunction with available prediction tools or other target databases, a high confidence dataset of global miRNA-mRNA interactions can be generated. We also describe further methods to prioritize identified interactions by integrating with mutation, copy number variation, methylation, or survival data to support observations and provide context. Finally, we discuss methods to experimentally validate selected novel targets.