Integrating prior knowledge in multiple testing under dependence with applications to detecting differential DNA...

Area of Science:

• Epigenetics
• Genomics
• Statistical Bioinformatics

Background:

• DNA methylation is a key epigenetic marker.
• Profiling DNA methylation uses platforms like tiling arrays and next-generation sequencing.
• DNA methylation data exhibits spatial correlation and varying CpG density, offering unique analytical insights.

Purpose of the Study:

• To introduce a robust testing and probe ranking procedure for differential DNA methylation detection.
• To incorporate features like spatial correlation and CpG density into a nonhomogeneous hidden Markov model.
• To improve upon existing methods for identifying differential methylation sites.

Main Methods:

• Development of a nonhomogeneous hidden Markov model for DNA methylation analysis.
• Application of a nonparametric symmetric distribution for two-sided hypothesis testing.
• Revisiting and extending the work of Sun and Cai (2009).

Main Results:

• The proposed model enhances probe ranking accuracy for differential methylation.
• The method demonstrates robustness against model misspecification through simulations.
• The framework shows improved operating characteristics compared to standard methods in real data analysis.

Conclusions:

• The novel hidden Markov model framework provides a powerful tool for differential DNA methylation analysis.
• The approach effectively utilizes inherent data characteristics for improved detection of methylation changes.
• This method offers a robust and high-performing alternative for identifying differential methylation sites.