Hi-C Pipeline - Usage Guide

Overview

This workflow analyzes Hi-C chromosome conformation capture data to identify compartments, TADs, and chromatin loops.

Prerequisites

conda install -c bioconda bwa-mem2 pairtools cooler
pip install cooltools

Quick Start

Tell your AI agent what you want to do: - "Analyze my Hi-C data for TADs and loops" - "Find A/B compartments in my Hi-C matrix" - "Process Hi-C FASTQ to contact matrix"

Example Prompts

Processing

"Generate a contact matrix from my Hi-C pairs"

"Balance my cooler file with ICE"

Analysis

"Call TADs using insulation score"

"Detect chromatin loops"

"Find compartment boundaries"

Input Requirements

What the Workflow Does

1. Alignment - Map Hi-C read pairs with bwa-mem2 -SP5M (mates aligned independently)
2. Pairs - Classify, filter, and deduplicate; judge library quality from long-range cis
3. Matrix - Build a cooler and zoomify to a multi-resolution .mcool
4. Balance - ICE normalization (required before any analysis)
5. Compartments - A/B eigenvector at 100kb, sign-phased by GC
6. TADs - Insulation-score boundaries across a window sweep at 10kb
7. Loops - Dot calling at 10kb IF the map is deep enough; else APA on known anchors

Tips

• Depth dictates the feature: compartments are cheap, TAD boundaries need a moderate map, de-novo loops need billions of contacts (~5B in Rao 2014). Decide the resolution from the depth, not the other way around.
• Resolution to feature: 100kb-1Mb for compartments, 10-40kb for TADs, 5-10kb for loops (1-2kb for Micro-C).
• Phasing is not optional: the compartment eigenvector sign is arbitrary until oriented by a GC or gene-density track.
• QC: the long-range cis (>=20kb) fraction is the one-number library readout; trans is a noise floor whose threshold is genome-size-dependent. High duplicate rate = low complexity (not fixable by sequencing deeper).
• Protein-directed assays branch out: HiChIP, PLAC-seq, and Capture Hi-C need FitHiChIP/MAPS/CHiCAGO (hi-c-analysis/hichip-plac-loops), not generic dot calling.