Summary

Hi-C library preparation involves a multi-step enzymatic protocol — crosslinking, restriction digestion, ligation, and library construction — where each step introduces specific failure modes that can compromise the final contact map. Unlike standard sequencing libraries, Hi-C library problems often do not appear until after sequencing, when the contact map reveals crosslinking artifacts, digestion failure, or excessive noise. This guide covers the most common Hi-C sample preparation risks, how to detect them early, and how to prevent or resolve them before committing to deep sequencing. If you are planning a Hi-C project, contact CD Genomics for consultation on sample preparation and library QC.

Key Takeaways

• Crosslinking time and temperature must be optimized for each sample type; both over- and under-crosslinking produce characteristic artifacts on contact maps
• Incomplete restriction digestion is the most common cause of low valid interaction ratios and should be checked by gel electrophoresis before ligation
• High duplicate rates (>40%) indicate low library complexity and usually trace back to over-crosslinking, insufficient input, or excessive PCR
• Pilot sequencing (1–3 million read pairs) before deep sequencing identifies library problems early
• The cis/trans ratio, P(s) decay curve, and fraction of valid pairs are the three most informative QC metrics for Hi-C libraries

Figure 1. Hi-C library preparation workflow with four critical risk control checkpoints — crosslinking, restriction digestion, proximity ligation, and library QC before sequencing.

What Good Hi-C Data Looks Like

Before troubleshooting problems, it helps to know what a passing Hi-C library looks like in the QC metrics.

A well-prepared Hi-C library from a mammalian genome should show the following characteristics in pilot sequencing:

These metrics are generated by standard Hi-C processing pipelines — HiC-Pro, HiCUP, Juicer, or HiCExplorer — and should be reviewed after pilot sequencing of 1–3 million read pairs before committing to full-scale sequencing.

Crosslinking That Fails

Crosslinking is the first chemical step in Hi-C, and it sets the ceiling for data quality. Both insufficient and excessive crosslinking produce distinct problems.

Over-crosslinking occurs when cells are exposed to formaldehyde for too long (beyond 15–20 minutes) or at too high a concentration (above 2%). The excessive crosslinks prevent restriction enzymes from accessing their recognition sites, reducing digestion efficiency. On the contact map, over-crosslinking appears as a skew toward very short-range interactions with weakened compartment and TAD signals. Duplicate rates also tend to be elevated because fewer unique DNA fragments are available for library preparation.

Under-crosslinking happens when crosslinking time is too short (under 5 minutes) or the formaldehyde concentration is too low (below 0.5%). Without sufficient crosslinking, spatially distant DNA fragments can ligate to each other during the ligation step, creating false “random” contacts. The contact map shows a weak diagonal, depressed long-range cis signal, and an elevated trans fraction.

The standard starting condition for mammalian cells is 1% formaldehyde at room temperature for 10 minutes, quenched with 0.125 M glycine. But this is a starting point, not a universal recipe:

Testing crosslinking efficiency on a small aliquot of sample before proceeding with the full library is a practical risk-reduction step.

Digestion That Stalls

Restriction enzyme digestion converts intact chromatin into fragments with compatible ends for proximity ligation. Incomplete digestion is the single most common cause of low valid interaction ratios in Hi-C libraries.

When digestion fails, the library contains a high proportion of:

• Same-fragment reads — both reads of a pair map to the same restriction fragment, indicating the fragment was not cut
• Dangling ends — reads that start at a restriction site but have no ligation partner
• Self-circles — a single fragment circularized by ligation

Together, these represent uninformative reads that reduce the effective sequencing yield.

Detecting digestion failure before sequencing. The digest can be checked directly by gel electrophoresis before the ligation step. After restriction digestion and heat inactivation, run 200–500 ng of digested DNA on a 1.5% agarose gel. A successful digest produces a smear centered at the expected fragment size for the chosen restriction enzyme (e.g., 100–500 bp for a 4-cutter like MboI or DpnII). A band at the high-molecular-weight region of the gel indicates incomplete digestion.

Common causes of incomplete digestion:

• Over-crosslinking — excess formaldehyde crosslinks block restriction sites
• Insufficient enzyme — the enzyme concentration or digestion time is too low for the amount of starting material
• Inhibitors — residual SDS, EDTA, or phenol from crosslinking or harvesting can inhibit restriction enzymes
• Sample-specific chromatin structure — heterochromatin regions are less accessible to restriction enzymes than open chromatin

Fixes. Increase digestion time from 2 hours to 4 hours or overnight. Add fresh enzyme after 2 hours for the overnight digest. Reduce crosslinking time if over-crosslinking is suspected. Include a no-ligation control to check whether the digest was complete.

Ligation Producing Noise

Proximity ligation joins the digested ends of DNA fragments that are physically close in the nucleus. When it works correctly, ligation produces valid long-range cis and trans contacts. When it goes wrong, it generates artifacts that waste sequencing depth.

Three types of ligation artifacts are common in Hi-C libraries:

Self-circles form when the two ends of a single restriction fragment ligate to each other. Self-circles appear in the QC report as read pairs mapped to the same fragment in an inward-facing orientation. A high self-circle rate indicates that the fragment ends were in close proximity but did not find a legitimate ligation partner — often because the digestion generated very short fragments that are prone to circularization.

Dangling ends occur when a restriction site overhang has been repaired and had adapters ligated but was never joined to another fragment. High dangling-end proportions suggest that ligation conditions were suboptimal — inefficient ligase, insufficient ligation time, or unfavorable buffer conditions.

Random ligation happens when fragments from distant genomic regions or different chromosomes ligate without being crosslinked. This produces trans contacts that look like real inter-chromosomal interactions. Random ligation is more likely when the ligation reaction is allowed to proceed for too long (overnight ligation without monitoring) or when the DNA concentration is too high, increasing the probability of random encounters.

The ligation step is sensitive to the DNA concentration. Most standard Hi-C protocols specify a dilution step before ligation to reduce the effective DNA concentration, favoring intra-molecular ligation over inter-molecular ligation. Skipping or altering this dilution is a frequent cause of excessive noise in the contact map.

Libraries with High Duplication

Duplicate reads — PCR or optical duplicates — reduce the effective sequencing depth and indicate that the library complexity is lower than it should be.

A duplication rate below 20% in pilot sequencing is a sign of a healthy Hi-C library. Rates above 40% at deep sequencing depth should trigger investigation.

Three common causes of high duplication:

• Low input material. Hi-C libraries require millions of cells as starting material (typically 1–10 million mammalian cells). Using fewer cells reduces the number of unique genome equivalents in the library, and the same molecules are sequenced multiple times regardless of sequencing depth.
• Over-crosslinking. Excess crosslinks reduce the fraction of chromatin that is accessible for digestion and ligation, shrinking the effective library complexity.
• Excessive PCR cycles. Over-amplification inflates the duplication rate. Most Hi-C libraries require 8–12 PCR cycles. If the library yield after 12 cycles is insufficient for sequencing, the problem is in an earlier step (crosslinking, digestion, ligation, or size selection), not the PCR.

Fixes include using more input material if available, reducing crosslinking time, and staying within the recommended PCR cycle range. If the input material is truly limited, consider a low-input Hi-C protocol rather than forcing a standard protocol with fewer cells.

Contact Maps That Stay Sparse

Even when QC metrics look good, the final contact map may lack the resolution needed to answer the biological question. This is a planning problem rather than a preparation failure, but it is the most common source of post-sequencing disappointment.

The effective resolution of a Hi-C contact map depends on sequencing depth, not library quality alone. A library with excellent QC metrics will still produce a sparse 1 kb-resolution map if sequenced to only 100 million read pairs.

The relationship between read depth and resolution is approximately:

• 500 million to 1 billion read pairs — 5–10 kb resolution, sufficient for compartment and TAD analysis in mammalian genomes
• 2–5 billion read pairs — 1–5 kb resolution, required for loop detection
• 10+ billion read pairs — sub-kb resolution, needed for fine promoter-enhancer analysis

For pilot studies or samples where the expected interaction frequency is low (e.g., cell-type-specific loops, rare cell populations), deeper sequencing is required to achieve the same effective resolution. The HiCRes tool can model the relationship between depth and resolution for a specific genome size and expected contact frequency.

When the contact map is too sparse for the intended analysis, the options are deeper sequencing of the existing library (if the library has sufficient complexity to support additional sequencing) or preparing a new library with more input material.

QC Checklist Before Submitting

Before submitting Hi-C libraries for full-scale sequencing, confirm the following:

• Crosslinking: Titrated for sample type; not over- or under-crosslinked
• Digestion: Confirmed by gel electrophoresis; >90% digested
• Ligation: Dilution step performed; ligation time appropriate
• Library QC: Bioanalyzer trace shows expected fragment distribution around 300–600 bp
• Pilot sequencing: ≥1 million read pairs; valid pair ratio ≥50%; duplicate rate ≤20%; cis/trans ratio in expected range
• P(s) curve: Monotonic decay without shoulder or plateau

Libraries that pass these checks have a high probability of producing high-quality contact maps. Libraries that fail any of these checks should be diagnosed and re-prepped rather than sequenced in the hope that the data will be usable — rescue from failed Hi-C libraries is rarely successful.

FAQ

1) How many cells are needed for a standard Hi-C library?

Standard Hi-C protocols recommend 1–10 million mammalian cells as starting material. Low-input Hi-C protocols can work with as few as 50,000 cells, but these use modified procedures for crosslinking, digestion, and ligation. Using fewer cells than the standard recommendation without protocol adjustments is a common cause of low-complexity libraries.

2) What is the most important QC metric for Hi-C library quality?

The fraction of valid interaction pairs — reads that represent genuine proximity ligation events — is the single most informative metric. It integrates the effects of crosslinking efficiency, digestion completeness, and ligation success. A library with ≥50% valid pairs across all QC pipelines (HiC-Pro, HiCUP, Juicer) is generally good.

3) Can I sequence a Hi-C library that fails QC checks?

Sequencing a library that fails QC is rarely cost-effective. A library with low valid pair ratio, high duplication, or poor P(s) decay produces a contact map that requires 2–5× more sequencing to reach the same effective depth as a passing library. In many cases, the failed library will never produce a usable contact map regardless of sequencing depth.

4) How do I distinguish under-crosslinking from random ligation?

Both conditions produce excessive trans contacts, but the P(s) decay curve distinguishes them. Under-crosslinking produces a shallow P(s) curve with elevated noise at all distances. Random ligation from improper ligation conditions tends to produce a characteristic shoulder or plateau in the short-range region of the P(s) curve before the expected decay.

5) What is a reasonable duplicate rate for a deep-sequenced Hi-C library?

A duplicate rate below 20% in pilot sequencing (1–3 million read pairs) is ideal. At deep sequencing depth (500 million+ read pairs), rates up to 40% are acceptable. Rates above 40% at any sequencing depth indicate low library complexity and should be investigated.

Related CD Genomics Services

• Hi-C Service (Chromosome Conformation Capture, 3C) — Genome-wide 3D chromatin architecture analysis
• Capture Hi-C Service — Targeted high-resolution chromatin interaction mapping
• Hi-C Service — Protein-centric chromatin interaction profiling
• Epigenomic Data Analysis — Bioinformatics for Hi-C and epigenomic data

References

1. Yardımcı GG, Noble WS. "Software tools for visualizing Hi-C data." Genome Biology. 2019;20:26. doi:10.1186/s13059-019-1662-4
2. Lajoie BR, Dekker J, Kaplan N. "The Hitchhiker’s guide to Hi-C analysis: practical guidelines." Methods. 2015;72:65-75. doi:10.1016/j.ymeth.2014.10.031
3. Servant N, Varoquaux N, Lajoie BR, et al. "HiC-Pro: an optimized and flexible pipeline for Hi-C data processing." Genome Biology. 2015;16:259. doi:10.1186/s13059-015-0831-x
4. DeMaere MZ, Darling AE. "qc3C: reference-free quality control for Hi-C sequencing data." PLOS Computational Biology. 2021;17(9):e1008839. doi:10.1371/journal.pcbi.1008839

Services mentioned in this article are provided for research use only and are not intended for clinical diagnosis, treatment, or personal health assessment.