Related Experiment Video
Updated: Mar 26, 2026
12:39
A Novel Bayesian Change-point Algorithm for Genome-wide Analysis of Diverse ChIPseq Data Types
Published on: December 10, 2012
11.7K
A full Bayesian partition model for identifying hypo- and hyper-methylated loci from single nucleotide resolution
Henan Wang1, Chong He2, Garima Kushwaha3
1Department of Statistics, University of Missouri, Columbia, Missouri, USA. hwg58@mail.missouri.edu.
BMC Bioinformatics
|January 29, 2016
Summary
This study introduces a Bayesian model to identify DNA methylation differences between sample groups, even with limited replicates. The model accurately detects differential methylation, distinguishing between hypo- and hyper-methylation in a single step.
Area of Science:
- Epigenetics
- Genomics
- Bioinformatics
Background:
- DNA methylation is a key epigenetic regulator of gene expression.
- Bisulfite sequencing provides single CpG resolution DNA methylation data.
- Limited biological replicates in sequencing studies pose statistical challenges for differential methylation analysis.
Purpose of the Study:
- To develop a statistical method for detecting differential DNA methylation loci.
- To address challenges posed by low sample sizes in methylation studies.
- To differentiate between hypo- and hyper-methylation.
Main Methods:
- A full Bayesian partition model was developed.
- The model leverages Bayesian principles to enhance statistical power for high-dimensional, low-sample-size data.
- The approach allows for simultaneous classification of loci as equal-, hypo-, or hyper-methylated.
Main Results:
- The proposed Bayesian model accurately identifies differentially methylated loci, particularly in low-coverage data.
- The model provides a one-step output distinguishing between equal, hypo-, and hyper-methylation.
- The R package MethyBayes implements the developed model.
Conclusions:
- The Bayesian partition model demonstrates superior performance compared to existing methods.
- The model achieves high statistical power while maintaining a low false discovery rate.
- Simulation studies and real data analysis validate the model's effectiveness.
