Differential analysis of RNA methylation with a novel statistical test

Chemical modification of mRNA ubiquitously exhibits in biological cells. As a new affinity-based sequencing approach, the methylated RNA immuno-precipitation sequencing (MeRIP-Seq) technology has greatly fueled study for mRNA methylation. However, conventional tests for contingency tables such as Fisher's exact test do not address the differ-ence in sequencing depth and thus cannot be applied for RNA methylation analysis. Here, we proposed a permutation based approach for detecting differentially methylated RNA from MeRIP-Seq data. The performance of this test was evaluated by the Type I error rate at different significance levels in simulated data. Additionally, the proposed test can effectively quantify the statistical significance of the interactive effect between RNA methylated sites and biological conditions. Moreover, we investigated the statistical power of the algorithm through simulation and compared it with alternative approaches that can be used for testing differential RNA methylation, including Woolf test, Breslow-Day test, bltest, and rhtest. Ultimately, the proposed test together with the R code showed promising performance. Mean-while, the proposed method is widely applicable to other scenarios, where homogeneity of association requires to be tested.