Abstract
Motivation: m6A methylation is a highly prevalent post-transcriptional modification in eukaryotes. MeRIP-seq or m6A-seq, which comprises immunoprecipitation of methylation fragments, is the most common method for measuring methylation signals. Existing computational tools for analyzing MeRIP-seq data sets and identifying differentially methylated genes/regions are not most optimal. They either ignore the sparsity or dependence structure of the methylation signals within a gene/region. Modeling the methylation signals using univariate distributions could also lead to high type I error rates and low sensitivity. In this paper, we propose using mean vector testing (MVT) procedures for testing differential methylation of RNA at the gene level. MVTs use a distribution-free test statistic with proven ability to control type I error even for extremely small sample sizes. We performed a comprehensive simulation study comparing the MVTs to existing MeRIP-seq data analysis tools. Comparative analysis of existing MeRIP-seq data sets is presented to illustrate the advantage of using MVTs. Results: Mean vector testing procedures are observed to control type I error rate and achieve high power for detecting differential RNA methylation using m6A-seq data. Results from two data sets indicate that the genes detected identified as having different m6A methylation patterns have high functional relevance to the study conditions. Availability: The dimer software package for differential RNA methylation analysis is freely available at https://github.com/ouyang-lab/DIMER.
Original language | English (US) |
---|---|
Article number | bbab309 |
Journal | Briefings in Bioinformatics |
Volume | 23 |
Issue number | 1 |
DOIs | |
State | Published - Jan 1 2022 |
Keywords
- RNA methylation
- differential analysis
- statistical methods
ASJC Scopus subject areas
- Information Systems
- Molecular Biology