HEPeak: An HMM-based exome peak-finding package for RNA epigenome sequencing data

Background: Methylated RNA Immunoprecipatation combined with RNA sequencing (MeRIP-seq) is revolutionizing the de novo study of RNA epigenomics at a higher resolution. However, this new technology poses unique bioinformatics problems that call for novel and sophisticated statistical computational solutions, aiming at identifying and characterizing transcriptome-wide methyltranscriptome. Results: We developed HEP, a Hidden Markov Model (HMM)-based Exome Peak-finding algorithm for predicting transcriptome methylation sites using MeRIP-seq data. In contrast to exomePeak, our previously developed MeRIP-seq peak calling algorithm, HEPeak models the correlation between continuous bins in an m⁶A peak region and it is a model-based approach, which admits rigorous statistical inference. HEPeak was evaluated on a simulated MeRIP-seq dataset and achieved higher sensitivity and specificity than exomePeak. HEPeak was also applied to real MeRIP-seq datasets from human HEK293T cell line and mouse midbrain cells and was shown to be able to recapitulate known m⁶A distribution in transcripts and identify novel m⁶A sites in long non-coding RNAs. Conclusions: In this paper, a novel HMM-based peak calling algorithm, HEPeak, was developed for peak calling for MeRIP-seq data. HEPeak is written in R and is publicly available.