cpg-island-feature-classification

CpG Island Feature Classification

Summary

Classify differentially methylated bases and regions relative to CpG islands, their flanking shores, and gene annotation features (promoters, exons, introns) to determine the genomic context and functional relevance of methylation changes. This skill uses genomation's annotation functions to generate percentage overlap tables that stratify differential methylation by feature class.

When to use

You have a methylDiff object containing differentially methylated bases or regions from bisulfite sequencing, gene annotation in BED or similar format (RefSeq, Ensembl), and CpG island coordinate files, and need to understand what fraction of your differential methylation signal falls within promoters vs. exons vs. introns and whether it clusters in CpG islands or their flanking shores.

When NOT to use

• Your input is already a feature-annotated table or matrix (i.e., annotation has already been performed).
• You have only raw bisulfite sequencing reads (FASTQ) and have not yet called methylation; use methylation callers (Bismark, MethylDackel) first.
• Your differentially methylated regions are from a non-mammalian organism for which CpG island definitions do not apply or are not validated.

Inputs

• methylDiff object (output from methylKit::calculateDiffMeth())
• RefSeq or Ensembl gene annotation BED file
• CpG island coordinate BED file (e.g., cpgi.hg18.bed.txt)

Outputs

• Percentage overlap table: differentially methylated bases by gene part (promoter/exon/intron/intergenic)
• Percentage overlap table: differentially methylated bases by CpG island context (CpGi/shore)
• Summary statistics table matching vignette format and counts

How to apply

Load your methylDiff object and convert gene annotation (RefSeq, Ensembl) and CpG island BED files into GRanges objects using GenomicFeatures or genomation. Execute annotateWithGeneParts() to overlap differentially methylated bases with promoter, exon, intron, and intergenic regions, recording the percentage of bases in each category. Then execute annotateWithFeatureFlank() to annotate the same bases relative to CpG islands and their flanking shores (typically 2 kb on each side), capturing the percentage overlap for CpGi vs. shore contexts. Compile the resulting percentage overlap statistics into a summary table. The rationale is that promoter and island contexts are functionally distinct from intergenic and shore contexts; stratification reveals whether differential methylation is enriched in regulatory or structural genomic compartments.