Systematic identification of A-to-I editing associated regulators from multiple human cancers

A-to-I editing is the most common editing type in humans that is catalyzed by ADAR family members (ADARs), ADAR1 and ADAR2. Although millions of A-to-I editing sites have recently been discovered, the regulation mechanisms of the RNA editing process are still not clear. Herein, we developed a two-step logistic regression model to identify genes that are potentially involved in the RNA editing process in four human cancers. In the first step, we tested the association of each editing site with known enzymes. To validate the logistic regression model, we collected 10 genes with 168 editing sites from multiple published studies and obtained a nearly 100% validation rate. ADAR1 was identified as the enzyme associated with the majority of the A-to-I editing sites. Thus, ADAR1 was taken as a control gene in the second step to identify genes that have a stronger regulation effect on editing sites than ADAR1. Using our advanced method, we successfully found a set of genes that were significantly positively or negatively associated (PA or NA) with specific sets of RNA editing sites. 51 of these genes had been reported in at least one previous study. We highlighted two genes: 1), SRSF5, supported by three previous studies, and 2) MIR22HG, supported by one previous study and two of our cancer datasets. The PA and NA genes were cancer-specific but shared common pathways. Interestingly, the PA genes from kidney cancer were enriched for survival-associated genes while the NA genes were not, indicating that the PA genes may play more important roles in kidney cancer progression.