hierarchical-dendrogram-interpretation

hierarchical-dendrogram-interpretation

Summary

Interpret hierarchical clustering dendrograms from methylation profiles to assess sample relatedness and grouping patterns. This skill applies distance-based clustering to reveal whether samples cluster by expected biological groups (e.g., treatment vs. control) and infers methylation-based similarity relationships.

When to use

You have a methylBase object containing aligned methylation calls across multiple samples and need to verify whether samples cluster by expected experimental condition (e.g., test vs. control) or identify unexpected sample relationships. Use this skill when sample relationships and data quality must be confirmed before downstream differential methylation analysis.

When NOT to use

• Input is a single sample or fewer than two samples—dendrograms require at least two samples to show relationships.
• Samples have been pre-filtered to remove low-coverage bases and you need quantitative differential methylation statistics rather than exploratory sample clustering.
• You have already identified batch effects and corrected them; re-clustering may not reflect the corrected structure without re-running clusterSamples().

Inputs

• methylBase object (output of unite() function containing merged methylation calls for all samples at positions covered in all samples)

Outputs

• Dendrogram object showing hierarchical relationships among samples
• Visual dendrogram plot with branch structure and sample labels

How to apply

Load a methylBase object from unite() containing base-pair coverage and methylation percentages across samples. Apply clusterSamples() to perform hierarchical clustering using correlation distance with Ward linkage, which produces a dendrogram showing pairwise sample relationships. Interpret the dendrogram by examining branch heights (reflecting correlation distance) and cluster membership—samples that cluster together have similar methylation profiles. Validate the dendrogram structure against expected biological groupings (e.g., test1 and test2 grouping separately from ctrl1 and ctrl2). Inspect outliers or unexpected groupings as indicators of batch effects, sample swaps, or biological heterogeneity that may require sample-level filtering via filterByCoverage() before proceeding to differential methylation testing.