A novel algorithm for calling mRNA m6A peaks by modeling biological variances in MeRIP-seq data

Area of Science:

• * Molecular Biology
• * Bioinformatics
• * Genomics

Background:

• * N(6)-methyl-adenosine (m(6)A) is the most abundant mRNA modification.
• * Methylated RNA Immunoprecipitation Sequencing (MeRIP-seq) enables transcriptome-wide m(6)A profiling.
• * Existing peak calling algorithms like exomePeak have limitations in modeling MeRIP-seq data characteristics.

Purpose of the Study:

• * To develop a novel, robust peak calling method for accurate m(6)A site detection.
• * To address limitations of existing methods by modeling read count variances and dependencies.
• * To improve the performance and reliability of m(6)A site identification from MeRIP-seq data.

Main Methods:

• * Proposed MeTPeak, a graphical model-based peak caller for m(6)A site detection.
• * Incorporated a hierarchical Beta variable layer to model read count variances.
• * Utilized a Hidden Markov model to capture read dependencies across genomic regions.
• * Developed a constrained Newton's method with a log-barrier function for parameter optimization.

Main Results:

• * MeTPeak demonstrated significant improvements in detection performance and robustness compared to exomePeak.
• * Applied to simulated and real biological datasets, MeTPeak showed enhanced accuracy.
• * Validated on public MeRIP-seq datasets, MeTPeak recapitulated known m(6)A patterns, confirming its efficacy.

Conclusions:

• * MeTPeak offers a superior approach for transcriptome-wide m(6)A site detection using MeRIP-seq data.
• * The method's ability to model data complexities leads to more reliable m(6)A profiling.
• * MeTPeak provides a valuable tool for advancing the understanding of m(6)A regulatory functions.