epigenetic-sample-stratification

epigenetic-sample-stratification

Summary

Stratify and visualize biological samples by their methylation profiles using unsupervised clustering and principal component analysis on base-pair resolution methylation data. This skill reveals sample relationships and groupings driven by DNA methylation similarity, enabling detection of batch effects, tissue/phenotype separation, and quality assessment of bisulfite sequencing experiments.

When to use

After merging methylation call files into a unified methylBase object (covering all samples at common base positions), apply this skill to assess whether biological replicates cluster together, whether case/control or treatment groups separate as expected, and to identify potential sample contamination or mislabeling before proceeding to differential methylation analysis.

When NOT to use

• Input methylation files have not been merged to a common set of covered positions (use unite() first)
• Sample number is very small (< 3 samples total); clustering and PCA require sufficient replication to reveal meaningful structure
• Methylation data come from highly heterogeneous tissues or cell types where within-group heterogeneity dominates; the skill may show dispersed rather than informative clustering

Inputs

• methylBase object (unified methylation matrix across all samples)
• sample metadata or grouping information (phenotype/treatment assignments)

Outputs

• dendrogram object from hierarchical clustering
• scree plot showing variance explained by principal components
• PC1 vs PC2 scatter plot with sample labels/colors

How to apply

Load a methylBase object produced by unite() function from methRead output files. Apply clusterSamples() to perform hierarchical clustering on methylation profiles using correlation distance with Ward linkage, generating a dendrogram that reveals sample grouping. Then apply PCASamples() to compute principal components and generate a scree plot showing the proportion of variance explained by each PC. Extract and visualize PC1 and PC2 as a 2D scatter plot to assess sample separation in methylation space. Interpret the dendrogram branch distances and PC scatter plot positioning to evaluate whether biological replicates show high similarity and whether experimental groups separate as hypothesized. High within-group correlation and clear between-group separation indicate good data quality and expected biological structure.