Hi-C turns the 3D organisation of the genome into a measurable contact matrix, and everything downstream — compartments, TADs, loops — is an interpretation of that matrix under a specific normalisation. Most analytical disagreements in the field are normalisation disagreements wearing a biological costume.
Chromatin conformation capture, briefly
The chemistry is elegant: crosslink DNA-protein interactions, digest with a restriction enzyme, fill in and ligate proximity ligated junctions, then sequence the chimeric fragments. Contact frequency between two genomic bins scales with their spatial proximity in the nucleus. Hi-C is genome-wide but low-resolution per locus; capture Hi-C and Hi-C variants trade coverage for targeted resolution. The readout is always a square symmetric matrix (or a tensor with multiple replicates), and the first analytical truth is that raw counts are dominated by genomic distance and by bin-level biases.
Library construction decisions that leak into results
- Restriction enzyme choice (HindIII, MboI/DpnII) sets the theoretical resolution floor.
- Digestion efficiency and ligation efficiency determine the fraction of informative junctions.
- Bioreplicate number matters more than sequencing depth past a point — reproducibility is the limiting factor.
- Crosslink reversal, biotin pull-down, and PCR cycles each introduce biases that must be monitored in QC.
.hic files and the Juicer ecosystem
The .hic format stores multi-resolution contact matrices with precomputed normalisation vectors, enabling zoom-in and zoom-out without re-deriving the matrix. Juicer handles the pipeline from FASTQ to .hic, and Juicebox is the visual front end. Normalisation options — KR (Knight-Ruiz) or ICE — matter: KR balances iteratively and handles sparsity better at fine scales, while ICE enforces equal marginal sums. Always state which normalisation you used; mixing them between sample and control invalidates comparisons.
A/B compartments
Compartments emerge from the first principal component (eigenvector) of the distance-normalised contact matrix: the sign of the eigenvector separates A (active, gene-rich, earlier-replicating) from B (inactive, heterochromatic, later-replicating). The sign is arbitrary and must be oriented using an anchor such as GC content or gene density. Compartment switching between conditions is a coarse but robust signal, and it is far more reproducible than fine-scale loop changes. If your compartment calls flip wholesale between replicates, the problem is data quality, not biology.
Topologically associating domains (TADs)
TADs are contiguous regions of elevated self-interaction, bounded by insulating elements that are frequently CTCF- and cohesin-rich. Callers differ substantially: directionality index segmentation, HiCseg, insulation score, and TopDom each impose different assumptions. TAD boundaries are largely conserved across cell types; it is boundary strength and intra-TAD contacts that vary. Treat TAD calls as scale-dependent annotations, not as discrete biological objects with a unique ground truth.
Loop calling with HiCCUPS
HiCCUPS detects focal enrichments over a local background modelled as a sum of donut-shaped neighbourhoods, using four components and aggregate peak analysis for validation. Loops require high depth — typically billions of contacts — and strong anchors, and they should be reproduced in independent replicates before being reported. Because loop calling is depth-sensitive, a "condition-specific loop" may simply be a depth artifact; downsample before claiming novelty.
From contacts to function
The payoff is linking distal elements to promoters: enhancer-promoter loops, and by extension the regulatory consequences of GWAS variants residing in non-coding regions. But contact does not equal regulation — the majority of promoter contacts have no measurable effect on transcription. Use Hi-C as one line of evidence alongside chromatin accessibility, histone marks, and QTL data, never as a standalone causal claim.
References
- Lieberman-Aiden E, van Berkum NL, Williams L, et al. Comprehensive mapping of long-range interactions reveals folding principles of the human genome. Science. 2009;326(5950):289–293.
- van Berkum NL, Lieberman-Aiden E, Williams L, et al. Hi-C: a method to study the three-dimensional architecture of genomes. J Vis Exp. 2010;(39):1869.
- Durand NC, Shamim MS, Machol I, et al. Juicer provides a one-click system for analyzing loop-resolution Hi-C experiments. Cell Syst. 2016;3(1):95–98.
- Rao SSP, Huntley MH, Durand NC, et al. A 3D map of the human genome at kilobase resolution reveals principles of chromatin looping. Cell. 2014;159(7):1665–1680.
- Dixon JR, Selvaraj S, Yue F, et al. Topological domains in mammalian genomes identified by analysis of chromatin interactions. Nature. 2012;485(7398):376–380.
- Nora EP, Lajoie BR, Schulz EG, et al. Spatial partitioning of the regulatory landscape of the X-inactivation centre. Nature. 2012;485(7398):381–385.
- Imakaev M, Fudenberg G, McCord RP, et al. Iterative correction of Hi-C data reveals hallmarks of chromosome organization. Nat Methods. 2012;9(10):999–1003.
- Servant N, Varoquaux N, Lajoie BR, et al. HiC-Pro: an optimized and flexible pipeline for Hi-C data processing. Genome Biol. 2015;16:259.
- Schmitt AD, Hu M, Jung I, et al. A compendium of chromatin contact maps reveals spatially active regions in the human genome. Cell Rep. 2016;17(8):2042–2059.