A hierarchical model for clustering m(6)A methylation peaks in MeRIP-seq data

Area of Science:

• Epigenetics
• Transcriptomics
• Bioinformatics

Background:

• High-throughput sequencing technologies like Methylated RNA Immunoprecipitation combined with RNA sequencing (MeRIP-seq) offer global views of N(6)-Methyladenosine (m(6)A) on the transcriptome.
• There is a need for advanced computational tools to analyze MeRIP-seq data for deeper insights into mRNA methylation functions.

Purpose of the Study:

• To develop a novel algorithm and open-source R package (MeTCluster) for clustering m(6)A methylation peaks.
• To uncover potential types of m(6)A methylation patterns and their location-specific functions.

Main Methods:

• Developed a hierarchical graphical model to analyze read count variance and cluster methylation peaks.
• Employed rigorous statistical inference for parameter estimation and cluster detection.
• Utilized an open-source R package, MeTCluster, for data analysis.

Main Results:

• MeTCluster accurately characterizes methylation peak clusters on simulated and real MeRIP-seq data.
• A novel pattern was identified: methylation peaks with lower enrichment cluster at the 5' end of mRNAs and lncRNAs.
• Peaks with higher enrichment are more frequently found in coding sequences (CDS) and at the 3' end of mRNAs and lncRNAs.

Conclusions:

• A novel algorithm based on a hierarchical graphical model was developed for clustering MeRIP-seq methylation peaks.
• The R package MeTCluster is publicly available for analyzing MeRIP-seq data.
• Results suggest that m(6)A functions may be location-specific within RNA molecules.