• Home
• Resource
• Choosing Hi-C as the Starting Point for 3D Genome Studies

Why teams often start with Hi-C (and what "starting point" really means)

Most 3D genome studies don't begin with a clean, assay-specific hypothesis. They begin with a biological observation:

• a perturbation changes transcription,
• a cell state shift suggests regulatory rewiring,
• a variant sits in a non-coding region,
• or a structural change (translocation, inversion, aneuploidy) raises mechanistic questions.

Early in that process, teams often don't know what kind of 3D signal they're dealing with.

Is it genome-wide (compartment shifts)? Domain-scale (TAD boundary changes)? Or a small number of protein-anchored loops at specific elements? A good first assay should help you classify the scope before you spend budget on a specialized design.

That is the core reason Hi-C is often the starting point: it provides a genome-wide baseline that can either (1) answer the first question directly, or (2) tell you what your second assay should be.

In practice: Teams sometimes request a highly specialized assay too early. When the effect later turns out to be broad rather than locus-limited, a genome-wide Hi-C baseline would have answered the question more directly—and with less redesign.

If your team wants a wider method-selection framework, CD Genomics has a decision guide comparing Hi-C, Micro-C, Capture Hi-C, and HiChIP.

What Hi-C is best at answering

Hi-C was introduced as a genome-wide chromosome conformation capture method by Lieberman-Aiden and colleagues in Comprehensive mapping of long-range interactions reveals folding principles of the human genome (Science, 2009). In modern projects, it's strongest when you need global organization more than fine local detail.

1) Genome-wide chromatin architecture and robust cross-condition comparisons

Hi-C produces a whole-genome contact map—an interaction matrix that summarizes 3D proximity across the genome. That makes it well suited to questions where the comparison is the deliverable:

• treated vs untreated,
• WT vs KO,
• responder vs non-responder models,
• longitudinal time courses,
• or multi-site cohorts where you need standardized outputs.

Because the view is genome-wide, you can detect broad architectural shifts without betting upfront on a handful of loci.

2) Compartments: a first-pass readout for global reorganization

A/B compartments are large-scale interaction patterns that often correlate with active vs inactive chromatin environments. When your question is "did the perturbation reorganize the genome broadly?", compartments are often one of the most stable, reviewable readouts.

3) TADs and domain-scale organization

Topologically associating domains (TADs) were formalized in mammalian Hi-C analyses by Dixon and colleagues in Topological domains in mammalian genomes identified by analysis of chromatin interactions (Nature, 2012). Domain-scale organization can be stable across many conditions—but when it changes, it can provide a structural explanation for regulatory rewiring.

A baseline domain map is also a useful coordination artifact: wet lab, bioinformatics, and project leadership can all point to the same structural units when discussing follow-up experiments.

4) A baseline map for multi-omics integration

Hi-C often becomes a scaffold for interpreting RNA-seq, ATAC-seq, and ChIP-seq. Even if the end goal is mechanistic (for example, enhancer-to-gene assignment), a genome-wide baseline can prevent over-interpreting local loop candidates that sit inside larger compartment or domain changes.

Key Takeaway: Hi-C is strongest when you need a whole-genome baseline for compartments, domains, and broad comparisons—not when the project depends on resolving a specific class of fine local loops from day one.

Where Hi-C can fall short (and what that implies for assay selection)

Hi-C is valuable because it is broad and standardized. But those same traits define where it is not the best first move.

Hi-C is not the default solution for loop-first questions

Many teams equate "3D genome" with "enhancer–promoter loops." Hi-C can nominate loop candidates, and in situ Hi-C can reveal looping principles (for example, Rao and colleagues' kilobase-resolution map in Cell (2014)).

But loop-level conclusions are also where Hi-C becomes sensitive to:

• sequencing depth and library complexity,
• replicate concordance,
• and the difference between plotting small bins and defending biological resolution.

If loop calls are the primary success metric, it is often more honest to treat Hi-C as a baseline and plan a second phase—unless you can truly support the depth and replicate strategy required.

Hi-C is not protein-centered

If the interpretability is defined by a protein factor or histone mark, a protein-anchored assay can be more direct.

HiChIP enriches contacts associated with a protein/mark of interest. That makes it a better starting assay when "contacts anchored by factor X" is the mechanism you need to test.

Hi-C is not locus-focused

If you only care about a predefined set of loci (GWAS regions, a promoter panel, a small gene set), genome-wide discovery may not be worth the sequencing burden.

Capture Hi-C concentrates power at selected loci, which can increase statistical support per dollar in those regions. The trade-off is that success becomes heavily dependent on capture design and panel consistency.

Hi-C does not provide single-molecule, multi-way contact context

Short-read Hi-C summarizes pairwise contacts across a population. If multi-way contacts and single-molecule structure are central to interpretation, a long-read 3D approach may add information Hi-C cannot.

When Hi-C is the right first assay—and when it is not

The fastest way to decide is to separate scope from mechanism.

Use Hi-C first when…

• You need a genome-wide baseline for compartments and domains.
• The effect may be broad or unknown, and you want to avoid over-designing the first experiment.
• Cross-condition comparison matters, and you need a defensible architectural map that collaborators and reviewers can inspect.
• You want to preserve optionality: run one baseline assay, then decide whether Micro-C, HiChIP, Capture Hi-C, or a long-read method is worth the extra complexity.

Don't start with Hi-C alone when…

• The central hypothesis is protein anchored (a factor/mark defines the loops you care about).
• You only care about a small predefined locus set and genome-wide discovery adds little.
• The main deliverable is ultra-fine short-range loop resolution, and you can't support the higher depth/QC burden.
• Multi-way contacts are a hard requirement.

A compact way to operationalize the decision is to ask: what would make the first dataset "successful" in a project review?

• If success is a publishable genome-wide baseline you can compare across conditions, Hi-C is usually a rational start.
• If success is a factor-anchored regulatory wiring map, a protein-anchored method may be the better first dataset.
• If success is high power at a predefined locus set, targeted capture may be the better first dataset.

Planning a Hi-C sequencing study around real constraints: samples, depth, and budget

If you're planning a study and want to align sample requirements, depth, QC, and deliverables upfront, you can start with Explore CD Genomics' Hi-C sequencing service overview.

Hi-C projects fail less often because of the mathematics and more often because of misalignment: a dataset is generated, but it cannot support the decision the team hoped it would.

Start from the decision the dataset must support

Before you talk about bin sizes, write one sentence:

• What decision will this dataset support in 6–10 weeks?

Examples:

• "Do we see compartment shifts across conditions?"
• "Do domain-scale changes justify a targeted follow-up?"
• "Is there enough signal to justify deeper sequencing for loops?"

That one sentence determines whether you should buy depth for compartments/domains—or whether you are implicitly trying to buy loop-level certainty.

Sample realities that drive interpretability

Sample quality and consistency often matter as much as sequencing.

This is where Hi-C sample requirements become practical, not theoretical: viability, consistent fixation, and avoiding sample degradation often determine whether the dataset is interpretable at the feature scale you care about.

From a planning standpoint, teams tend to underestimate:

• heterogeneity (mixed populations dilute signal),
• variability in fixation and digestion behavior,
• and how much precious input constrains replicate strategy.

Those constraints don't necessarily mean "don't do Hi-C." They mean "be explicit about what the dataset can reasonably support."

Hi-C sequencing depth should be tied to the call set

A bin size is not a biological deliverable. Resolution is a claim about interpretability.

A defensible order of operations is:

1. define the feature class (compartments, domains, loops),
2. define the comparison (single sample vs differential),
3. agree on acceptance criteria (QC + replicate concordance),
4. then choose Hi-C sequencing depth.

In practice: A common planning mistake is to ask for loop-level conclusions before the team has aligned on whether domain-scale changes would already answer the project question. This usually creates unnecessary pressure on sequencing depth, budget, and expectations.

Budget-honest designs reduce rework

If you have 10–30 samples across conditions, the "best" technical assay may not be the best project assay.

A phased plan is often more honest:

• Phase 1: generate a genome-wide baseline to classify the signal.
• Phase 2: escalate only where the baseline leaves a specific, interpretable gap.

That approach tends to reduce total cost because it avoids buying deep sequencing for every sample before you know what you are chasing.

What makes a Hi-C dataset actually usable

This is where many projects succeed or fail.

A dataset is "usable" when it supports the comparisons and calls you planned for, and when the delivery package makes those calls reviewable for both scientists and project leads. That is also the practical meaning of Hi-C analysis in a service context: not just producing matrices, but producing interpretable calls with clear QC and parameters.

A decision-ready set of Hi-C deliverables typically includes:

Reviewable matrices and browser-ready files

• matrices at multiple resolutions,
• normalized maps (with the method stated),
• and browser-ready files (commonly .hic and/or .cool).

Compartments and domains with explicit parameters

• compartment outputs that support condition-to-condition comparisons,
• domain/TAD outputs with boundary metrics,
• and clear reporting of parameters so results are reproducible.

Loop candidates (when loops are in scope)

If your project requires loop candidates, a usable output should include:

• loop candidates with statistical support,
• a replicate logic statement,
• and conservative language about what is and is not resolved.

A QC report with acceptance criteria

A useful QC report should make it obvious whether:

• the library behaves like a real Hi-C library,
• comparisons are defensible,
• and deeper sequencing is likely to improve interpretation at the chosen scale.

A common planning mistake is to treat a large dataset as automatically "decision-ready." It isn't—unless you defined the decision it must support.

Common planning mistakes teams make when starting with Hi-C

These are the failure modes that most often waste budget.

Mistake 1: Treating Hi-C as a magic answer for every chromatin question

Hi-C is a baseline assay. If you ask it to behave like a protein-centric loop mapper, you will either overpay for depth or end up underpowered.

Mistake 2: Choosing a resolution target before defining the biological decision

If you can't state what call set will be delivered (compartments, domains, loops), you can't rationally plan depth. A resolution label without a call set is a false promise.

Mistake 3: Underestimating sample quality constraints

Sequencing cannot fully compensate for poor or inconsistent input. When samples are heterogeneous or compromised, the right move is often to re-scope what conclusions are defendable—and to plan deliverables accordingly.

Mistake 4: Assuming more reads automatically fix interpretation

More reads help when you are limited by sparsity. If the limiting factor is library complexity, protocol bias, or sample heterogeneity, depth can become an expensive way to confirm the wrong assumption.

Mistake 5: Not defining deliverables and acceptance criteria upfront

If wet lab expects browser tracks and a clear answer, bioinformatics expects iterative exploration, and project management expects a milestone report, you need to align before sequencing—not after.

Why teams often begin with Hi-C before moving to a specialized assay

A good 3D genomics program looks more like a funnel than a single shot.

Hi-C serves as a screening layer for genome-wide chromatin architecture:

• If the baseline shows broad reorganization, you may stay at the compartment/domain level and integrate with other omics.
• If the baseline points to localized regulatory questions, you can escalate selectively.

That phased logic is usually more budget-honest and lower risk than forcing a specialized assay at day one.

For a general overview of 3D genome organization and why different feature scales exist (compartments vs domains vs loops), a commonly cited review is Bonev and Cavalli's Organization and function of the 3D genome (Nature Reviews Genetics, 2016).

Conclusion: start with the question, but let Hi-C give you the baseline

If your team needs a genome-wide structural starting point, Hi-C sequencing is often the most defensible first assay. It gives you a baseline architecture map for compartments and domains, supports broad comparisons, and helps you decide whether a specialized follow-up is truly necessary.

Next steps (RUO): If you want to see what an end-to-end workflow and delivery package can look like, explore CD Genomics' Hi-C sequencing service page. If you're still deciding between assays, start with the decision guide comparing Hi-C, Micro-C, Capture Hi-C, and HiChIP.

FAQ