contact-frequency-aggregation-by-genomic-feature

contact-frequency-aggregation-by-genomic-feature

Summary

Aggregate Hi-C contact frequencies around genomic features (e.g., CTCF binding sites) to identify local topological patterns and average interaction strengths. This skill extracts enriched contact neighborhoods and reveals how specific proteins or regulatory elements organize chromosome structure.

When to use

You have a cooler Hi-C contact matrix, a set of genomic features (e.g., CTCF peaks, enhancers, or TAD boundaries defined in BED format), and want to quantify average contact patterns around those features to detect local organization principles. Use this when investigating how specific architectural proteins or cis-regulatory elements shape three-dimensional genome folding.

When NOT to use

• Input Hi-C data has not been normalized for sequencing depth or bin-level biases; apply iterative correction or ICE normalization first.
• Genomic features are very sparse (<<100 sites per chromosome) or have extreme size variation; aggregation may produce unstable averages.
• You are interested in single-feature contact patterns rather than aggregate behavior; use direct contact extraction or focused visualization instead.

Inputs

• cooler Hi-C contact matrix file (.cool or .mcool)
• genomic feature coordinates (BED format or similar track defining feature locations)
• bin size and genome reference (implicit in cooler file)

Outputs

• 2D pileup matrix (aggregated contacts around features, typically N×N array)
• average contact frequency heatmap
• enrichment metric or fold-change relative to genome-wide contact distance curve

How to apply

Load a cooler file containing the binned Hi-C contact matrix and a BED or similar feature track defining genomic regions of interest. Extract or pre-compute the set of genomic coordinates for each feature (e.g., CTCF binding sites from ChIP-seq). Use cooltools' pileup or aggregation functions to stack contact matrices centered on each feature, normalizing by genomic distance to account for the distance-decay of contact frequency. Average the stacked matrices to produce a consensus 2D map showing how contacts are enriched or depleted relative to the feature. Evaluate the output by examining whether the resulting heatmap shows symmetry (expected around a central feature) and whether contact strength falls away from the feature center; compare against random or shuffled feature coordinates as a null model.