hi-c-normalization-and-bias-correction

Hi-C Normalization and Bias Correction

Summary

Normalize and correct systematic biases in Hi-C contact matrices to enable accurate comparison of interaction frequencies across genomic regions and between samples. This skill is essential for removing technical artifacts (e.g., restriction enzyme cutting efficiency, GC content, mappability) that distort raw contact counts and obscure true 3D chromatin structure.

When to use

Apply this skill after generating raw Hi-C contact matrices from aligned reads (post-merge, pre-analysis). Use it when you need to compare interaction frequencies between regions with different sequence properties, merge data across multiple replicates or experiments, or prepare matrices for downstream analysis (peak calling, domain detection, comparative genomics). Essential when raw contact counts show systematic variation correlated with genomic features rather than true interaction strength.

When NOT to use

• Input is already a normalized or published Hi-C matrix from a repository — re-normalization may introduce artifacts or is redundant.
• Data is from a non-standard restriction enzyme or in-situ protocol where Juicer's built-in bias model has not been validated — consider custom normalization pipelines.
• Sample size or sequencing depth is extremely low (< 1M valid pairs) — normalization may amplify noise rather than reveal signal.

Inputs

• Merged, deduplicated Hi-C alignment file (merged_nodups format or equivalent SAM/BAM)
• Reference genome sequence (FASTA)
• Restriction enzyme specification (e.g., HindIII, MboI)
• Chromosome sizes file (chrom.sizes)
• Juicer pipeline configuration (genome ID, queue parameters, computational resources)

Outputs

• Normalized Hi-C contact matrix (.hic file format)
• Bias correction factors (per-bin weights, embedded in .hic)
• Contact matrix statistics and quality metrics

How to apply

The Juicer platform includes normalization as part of its unified pipeline: after read alignment, deduplication, and chimera filtering, the pipeline constructs a raw contact matrix and applies normalization during .hic file creation. The normalization step removes biases by modeling how sequence-level and mappability factors affect observed contact counts. Configure the pipeline with the appropriate reference genome and restriction enzyme (e.g., HindIII, MboI) so that bias correction can account for restriction site distribution and local sequence properties. Execute the final pipeline stage to generate the normalized .hic output file. Verify that the resulting contact matrix shows biologically plausible patterns (e.g., strong diagonal, decay with genomic distance) and that inter-regional comparisons are no longer confounded by GC content or mappability differences.