Why Chromatin Contacts Matter
A candidate variant may lie in a regulatory element far from the gene it influences along the linear DNA sequence. Gene maps can suggest possible links, but they cannot show whether two genomic regions are brought into spatial proximity in a particular cell population. Chromatin-contact assays provide evidence about that proximity: they help identify which promoters contact a candidate regulatory region and how strong those measured contacts appear within the assay.
That evidence is useful, but its interpretation requires care. A measured contact does not by itself show that the variant changes the contact, that the contacted gene’s expression changes, or that the gene mediates the trait association. The most defensible interpretation is narrower: a contact supports a possible regulatory-element-to-promoter link in the assayed biological context.
What Contact Assays Measure
Many contact assays use chemical crosslinking to preserve some close spatial associations, fragment the DNA, ligate nearby fragments, and sequence the resulting junctions. The details vary across protocols. In Hi-C, ligation products are used to build a genome-wide contact map: genomic regions are arranged along both axes, and each map cell summarizes the observed contacts between a pair of regions. The map is commonly reported at a chosen bin size, such as 10 kilobases (kb). A larger value in a map cell generally means more captured contacts in that experiment after the stated processing, not a direct measurement of a fixed physical distance.
Promoter-capture Hi-C (PCHi-C) uses probes to enrich ligation products involving selected promoter regions. This makes it more focused on promoter contacts than genome-wide Hi-C, but it does not necessarily cover every promoter or every possible regulatory region equally. Capture-C and related approaches can focus on one or more selected viewpoints, such as a promoter, and measure contacts involving those regions. Micro-C uses a different fragmentation strategy and can provide finer-scale contact information under suitable experimental and sequencing conditions. These assays differ in design, coverage, and processing; their output values should not be treated as interchangeable measurements.
A contact map is therefore shaped by both biology and measurement. Sequencing depth, restriction or fragmentation pattern, probe design, binning, mapping, and normalization can all influence the observed signal. A contact score from one assay is not automatically comparable with the same-looking score from another assay, or with a score from a different cell sample.
Resolution, Strength, and Context
Resolution describes the scale at which the map can distinguish contacts between regions. It depends on the assay design, data depth, and analysis choices; a stated bin size is not a guarantee that every feature at that scale is measured precisely. If a candidate enhancer and several neighboring elements fall inside one bin, a contact between that bin and a promoter cannot identify which element is responsible. Even a fine-scale map may be noisy or incomplete.
Contact strength should be interpreted relative to an appropriate reference. Nearby regions often show more contacts simply because of their linear genomic proximity. Analysts may compare a contact with a distance-aware background, use a normalized enrichment, or examine it against other contacts in the same experiment. The comparison must be appropriate to the assay and processing. A raw count alone is usually insufficient: 30 reads may be compelling in a sparse assay and unremarkable in a deeply sequenced one.
Cell context matters because chromatin organization can vary across cell types, cell states, and experimental conditions. A contact detected in one cell sample is evidence for that sample, not automatic evidence that the same link operates in every tissue relevant to a trait. When interpreting a map, report which cell or sample was assayed and consider whether it is relevant to the biological question. A contact map from an unrelated cell context may still suggest a structural possibility, but it is weaker evidence for a context-specific regulatory mechanism.
| Feature to assess | Useful question | Interpretive caution |
|---|---|---|
| Resolution | Can the map distinguish the candidate element from nearby regions? | A bin-level contact may combine several elements. |
| Contact strength | Is the signal enriched relative to a suitable background in this assay? | Raw counts depend on depth and protocol. |
| Assay coverage | Were the relevant promoter and candidate region measured? | An unobserved contact may reflect limited coverage. |
| Cell context | Which cell population or condition produced the map? | A contact in one context need not occur in another. |
| Variant placement | Does the candidate variant lie in the contacted regulatory region? | A contact involving a broad region does not pinpoint a causal base. |
Worked Example: Using a Hi-C Contact to Compare Candidate Genes
Suppose a hypothetical fine-mapping analysis gives variant V1 a posterior inclusion probability of 0.70. V1 lies in a candidate enhancer at coordinates 1,204,900–1,206,100. A Hi-C map from a relevant cultured cell sample uses 10-kb bins. The enhancer-containing bin has a normalized contact enrichment of 2.4 with the promoter of Gene A and 1.1 with the promoter of Gene B. The promoters are 125 kb and 145 kb from the variant, respectively.
Step 1: Check whether the map can represent the proposed link. The candidate enhancer is 1.2 kb long, smaller than the 10-kb bin. The contact is therefore evidence about the bin containing the enhancer, not a precise measurement of the enhancer alone. The promoter regions are represented in the map, so both proposed contacts are observable at this map’s resolution.
Step 2: Compare the reported values within their proper scope. The reported enrichment for Gene A is 2.4, compared with 1.1 for Gene B. The ratio is \(2.4/1.1 \approx 2.18\), so the normalized enrichment with Gene A is about 2.18 times the enrichment with Gene B in this map. This is a comparison of assay-derived enrichments, not a statement that Gene A is 2.18 times as likely to be the target.
Step 3: State the conclusion with its limits. In this hypothetical map, the contact pattern gives stronger support to a possible link between the enhancer-containing region and Gene A than to Gene B. V1’s posterior inclusion probability concerns variant causality under the fine-mapping model; the contact enrichment concerns spatial proximity in this assay. Neither number establishes that V1 changes the enhancer or regulates Gene A. The bin’s width also prevents attributing the signal uniquely to the 1.2-kb enhancer.
Why Coarse Resolution Can Be Ambiguous
A useful way to see the resolution problem is to express a coarse-bin contact as the sum of contacts involving the smaller regions inside that bin. Let \(B\) be a coarse genomic bin containing several subregions, and let \(P\) be a promoter region. If the assay’s counts are aggregated across those subregions, then the bin-level count is
where \(E_1,\ldots,E_m\) are the subregions in \(B\), and \(C(E_j,P)\) is the count assigned to contacts between subregion \(E_j\) and promoter \(P\). This identity explains why a large coarse-bin count does not identify which \(E_j\) contributed the signal. It also shows why finer resolution is valuable only when the data support it: splitting counts across smaller regions can improve localization, but may leave too few observations for stable estimates.
For instance, suppose a hypothetical 50-kb bin contains four 10-kb subregions. Its total count with a promoter is 24, with respective contributions of 12, 4, 5, and 3. The first subregion contributes \(12/24 = 0.50\), or half of the total; the other half comes from the remaining regions. If a candidate variant lies in the first subregion, the coarse map supports a regional contact but cannot, by itself, establish that the candidate element accounts for all 24 counts. The finer breakdown is informative here because it preserves the subregion-level counts.
Worked Example: Interpreting a Promoter-Capture Signal
Consider a hypothetical PCHi-C experiment in a specified cell sample. The assay captures the promoter of Gene C. A candidate regulatory interval containing a fine-mapped variant has a normalized contact count of 36 with that promoter. For comparison, the analyst’s distance-matched background for comparable intervals in the same experiment has a mean count of 12.
Step 1: Calculate the relative signal. The interval’s count divided by the background mean is
Thus, the interval’s normalized count is three times the stated background mean. This comparison is interpretable because the example specifies a distance-matched background from the same experiment; it does not make the count a universal measure of contact strength.
Step 2: Check what the assay covers. The promoter of Gene C is a captured viewpoint, and the candidate interval has a reported contact with it. This is direct evidence in the assay for spatial association between the captured promoter and the interval. If the candidate interval is broad, the signal still does not show which base or subelement generated the association.
Step 3: Give a calibrated interpretation. The three-fold relative signal supports a possible interval-to-Gene C link in the assayed cell sample. It does not mean the variant has a three-fold effect on expression, and it does not show that Gene C mediates a trait association. The contact assay measures proximity-related ligation evidence, not the effect of changing the variant.
Combining Contact Evidence With Variant-to-Gene Reasoning
A contact is most informative when it addresses a specific proposed link. First establish that the candidate variant or its annotated regulatory element lies in the region represented by the contact. Then identify the contacted promoter and the assay’s effective resolution. Finally assess the strength and context of the signal. If the variant is merely somewhere in a broad contact bin, the link is less localized than if a well-resolved candidate element is specifically measured.
Contact evidence complements, rather than replaces, other functional annotation. A regulatory-element annotation can indicate that a region may have regulatory activity; a contact can indicate that it is spatially associated with a promoter. The two address different questions. Neither alone demonstrates that the candidate allele alters regulation. Contact evidence is particularly useful when it provides a plausible long-range connection that linear proximity would not suggest, but the nearest contacted promoter is not automatically the causal target.
Worked Example: Comparing Maps From Two Cell Samples
In a hypothetical study, the same promoter-capture design is applied to two cell samples. For a candidate enhancer interval, the normalized contact count with Gene D’s promoter is 28 in Sample 1 and 7 in Sample 2. The corresponding distance-matched background means are 14 and 7.
Step 1: Calculate enrichment within each sample. For Sample 1, the relative signal is \(28/14 = 2.00\). For Sample 2, it is \(7/7 = 1.00\). The interval is enriched two-fold over the stated background in Sample 1 and equals the background mean in Sample 2.
Step 2: Avoid comparing raw counts as if they were equivalent. The raw count 28 is four times 7, but the background levels also differ. The within-sample comparisons show that the evidence for enrichment is stronger in Sample 1. This does not prove that the contact exists only in Sample 1: differences in depth, capture performance, and sampling uncertainty may also contribute.
Step 3: Limit the biological conclusion. The data support a candidate enhancer-to-Gene D contact in Sample 1 more clearly than in Sample 2 under this assay. Interpretation for a trait should consider which sample is biologically relevant and whether the contact is reproducible. These numbers alone do not demonstrate a variant-dependent regulatory effect or establish a causal gene.
Common Mistakes and a Practical Review
- Mistake: “The contact is strong, so the variant regulates that gene.” A contact indicates assay-detected proximity, not a variant’s functional effect. The correction is to describe a supported possible link and seek separate evidence for allele-dependent regulation.
- Mistake: “The map has a contact for the bin, so the enhancer inside it is responsible.” A bin can contain several elements. Check whether the assay resolves the candidate interval or only a larger region.
- Mistake: “The largest raw count identifies the strongest biological link.” Counts depend on sequencing depth, capture design, distance, and processing. Use an appropriate normalization or comparison within the experiment, and report what was compared.
- Mistake: “A contact found in one cell type applies to all tissues.” Contact maps describe the assayed sample. State that context and avoid generalizing beyond the evidence.
- Mistake: “No observed contact means no regulatory link exists.” The relevant region may be poorly covered, the signal may be below detection, or the assay may lack resolution. Treat absence of a detected contact as limited evidence, not definitive exclusion.
Check Your Understanding
Answer each question by identifying what the contact evidence supports and what it leaves unresolved.
- Why can a contact between a 50-kb bin and a promoter fail to identify which enhancer inside the bin is involved?
- A candidate region has a normalized contact count of 18, and its same-assay distance-matched background mean is 6. What is the relative signal, and what does that calculation not establish?
- Why should raw contact counts from two different assay designs not be compared as though they were on one common scale?
- A promoter-capture assay detects a contact in one cell sample but not another. Name two reasons, other than a genuine biological difference, that could contribute to the observed contrast.
- What additional kind of evidence would be needed to claim that a candidate variant changes regulation of a contacted gene, rather than merely lying in a region that contacts its promoter?