FUSE: data-driven functional segmentation of DNA methylation data.

Area of Science:

• Genomics
• Epigenetics
• Bioinformatics

Background:

• DNA methylation (DNAm) patterns are crucial for gene regulation and chromatin structure.
• Identifying biologically relevant DNA methylation blocks from whole-genome bisulfite sequencing (WGBS) data is challenging.
• Current methods often use fixed genomic windows, not reflecting actual methylation patterns.

Purpose of the Study:

• To present FUSE, a data-driven segmentation method for defining DNA methylation blocks.
• To capture intrinsic methylation segments directly from WGBS data by analyzing multiple samples.
• To facilitate post hoc methylation analyses by aggregating coherent CpG sites.

Main Methods:

• FUSE is a data-driven segmentation method analyzing multiple WGBS samples.
• It identifies spatially homogeneous methylation blocks shared across a cohort.
• The method allows for different methylation states across samples.

Main Results:

• FUSE identified segments that significantly overlap with promoters, enhancers, and repetitive elements in 61 ENCODE WGBS samples.
• The method demonstrated high sensitivity in recovering true segment breakpoints in synthetic data, even with noise.
• FUSE successfully aggregated coherent CpG sites into candidate segments.

Conclusions:

• FUSE provides a robust approach to defining biologically meaningful DNA methylation blocks.
• This facilitates downstream analyses such as differential methylation testing.
• The method advances the understanding of DNA methylation patterns and their link to gene regulation.