From Chromatin Tracks to Evidence
A candidate variant can fall inside a region with strong chromatin accessibility or a histone-mark signal associated with regulatory activity. That overlap can make the region biologically interesting, but it answers only a regional question: what chromatin state was measured at this location in the assayed material? It does not, by itself, show that the candidate variant changes the state or causes the trait association.
This distinction is essential when interpreting regulatory annotations at a fine-mapped locus. An assay produces a signal from chromatin in a sample; analysis may summarize that signal as a peak or interval; comparison with a candidate-variant coordinate establishes overlap or non-overlap. Each step supports a different claim. Keeping the claims separate prevents a plausible regulatory annotation from becoming an unjustified claim of variant function.
Accessibility Tracks
ATAC-seq uses a transposase that preferentially inserts sequencing adapters into accessible DNA. DNase-seq uses DNase I cleavage, which is more frequent in accessible chromatin. After sequencing and processing, both assays can be displayed as a signal track across genomic coordinates. A local elevation in signal may be called a peak: an interval whose signal is judged elevated relative to a chosen background under a particular analysis procedure.
A peak is a useful summary, not a direct boundary between “open” and “closed” DNA. Peak locations and widths can depend on read depth, background estimation, processing choices, and the peak-calling threshold. The numerical signal in a track may be a normalized read count, an enrichment, or another assay-specific quantity. Its units and scale are not automatically comparable across assays, studies, or samples.
Accessibility is consistent with the possibility that regulatory proteins can access a region. It does not identify which protein is bound, establish that transcription is occurring, or reveal whether a particular nucleotide is responsible for the accessibility. A region can be accessible for reasons unrelated to the candidate variant under consideration.
Histone Marks as Evidence About Genomic State
Histone-mark assays measure enrichment of DNA associated with a specific histone modification. The marks are often informative about broad genomic states, but their interpretation is associative and context-dependent. Common examples include:
- H3K4me3 is commonly enriched near active promoters.
- H3K4me1 is often associated with enhancer-like regions, including regions that may not show strong evidence of current activity.
- H3K27ac is commonly associated with active promoters and enhancers.
- H3K27me3 is associated with Polycomb-mediated repression, while H3K9me3 is associated with some forms of compact, repressive chromatin.
These are interpretive tendencies, not deterministic labels. For example, H3K4me1 enrichment alone does not establish that a region is an active enhancer, and H3K27ac enrichment does not prove that a specific variant activates transcription. Several marks can occur in a region, and a signal can be affected by assay quality, background, sample composition, and the analysis pipeline. The measured signal is evidence about the region in the assayed material—not a direct reading of the molecular consequence of a single base change.
Three Levels of Interpretation
A disciplined annotation separates three questions. First, what did the assay measure? Second, what regional state is that pattern consistent with? Third, what does it establish about the candidate variant? The first two may be addressed by a track and its assay metadata. The third generally requires evidence that distinguishes the candidate variant or its alleles from the surrounding region.
| Observation | Reasonable interpretation | Conclusion not justified by this observation alone |
|---|---|---|
| ATAC-seq or DNase-seq peak spans a candidate position | The position lies in a region with accessibility signal in the assayed material. | The variant changes accessibility or is causal for the trait. |
| H3K27ac enrichment overlaps the position | The region has a mark commonly associated with active regulatory chromatin. | The variant increases expression of a particular gene. |
| H3K4me1 enrichment without strong H3K27ac | The region may have an enhancer-associated or less clearly active state. | The region is definitely an active enhancer, or the variant is functional. |
| No peak is reported at the position | The chosen analysis did not report a peak there. | The chromatin is certainly inaccessible in all relevant cells or conditions. |
Worked Examples
Worked Example: Checking Variant Overlap With an Accessibility Peak
Consider a hypothetical candidate variant at position 100,245 in a genome coordinate system. In an ATAC-seq track, a called peak spans positions 100,200 through 100,310. The normalized ATAC signal at the variant position is 8.4 track units; the local background signal is 2.1 units. Assume the position and peak are reported on the same reference assembly and use compatible coordinate conventions.
Overlap. The variant position is within the peak because \(100{,}200 \leq 100{,}245 \leq 100{,}310\). Thus, it overlaps an ATAC-seq peak in this sample.
Signal relative to background. The ratio of the displayed signal to local background is
The signal at the position is four times the stated local background on this track’s scale. This is a description of the supplied values, not a universal threshold for accessibility; the scale and peak-calling procedure matter.
Conclusion. The evidence supports saying that the candidate lies in an accessible region in the assayed material. It does not show that the variant causes the peak, alters accessibility, or explains the trait association. The assay measures a region-wide signal, and nearby DNA can contribute to it.
Worked Example: Interpreting a Mixed Histone-Mark Pattern
In a hypothetical assay, a candidate position lies within an interval with H3K4me1 signal of 5.2 normalized units and H3K27ac signal of 1.1 units. The stated local backgrounds are 1.3 units for H3K4me1 and 1.0 unit for H3K27ac. Assume each track has been processed against its own background and that the interval coordinates match the candidate’s reference assembly.
The signal-to-background ratios are \(5.2/1.3=4.0\) for H3K4me1 and \(1.1/1.0=1.1\) for H3K27ac. The first track therefore shows stronger enrichment relative to its stated background than the second. Because the two marks are measured on different assay scales, the raw values 5.2 and 1.1 should not be compared as if they shared a common unit.
H3K4me1 enrichment is consistent with an enhancer-associated region, while the comparatively weak H3K27ac signal provides less support for a strongly active state in this assay. This pattern does not establish that the interval is inactive: thresholds, sample composition, background, and assay sensitivity all matter. Nor does it establish that the variant has any functional effect. The defensible conclusion is that the interval has an H3K4me1-enriched pattern with little H3K27ac enrichment relative to the supplied background in this particular assay.
Worked Example: Why Regional Overlap Is Not Variant-Specific Proof
Suppose, purely for illustration, that before considering an accessibility result, a candidate variant has a 0.02 probability of being causal. Assume that a specified accessibility-peak overlap occurs for 80% of causal variants and for 25% of noncausal variants in a hypothetical, appropriately defined reference setting. These rates are illustrative and are not estimates for a real assay.
Let \(C\) denote “the candidate is causal” and \(O\) denote “the candidate overlaps the peak.” By Bayes’ rule,
Substituting the stated values gives
Under these illustrative assumptions, overlap changes the probability from 2% to about 6.13%, an increase that is informative but far from proof. The calculation depends on the assumed prior and on how often overlap occurs among causal and noncausal variants. In an actual analysis, those rates require relevant, well-calibrated evidence; one cannot infer them from the track alone. The example shows why a regional annotation can support a hypothesis without identifying the causal variant.
A Short Justification: Why Overlap Is a Limited Claim
An accessibility peak is an interval-level observation: it reports signal across a stretch of DNA. A candidate coordinate inside that interval satisfies a geometric relation, \( \text{position} \in \text{peak interval} \). That relation says nothing by itself about which base produced the signal. Multiple variants, or no causal variant at all, can lie in or near an accessible interval. Therefore, overlap is sufficient to describe the candidate as located in a region with accessibility signal, but it is not sufficient to infer a variant-specific molecular effect.
The same logic applies to histone marks. If an enriched interval contains the candidate position, the mark supports a description of the interval’s chromatin state in the assayed sample. It does not isolate the candidate’s contribution. Evidence that specifically tests alleles or changes the variant addresses a different question from a regional chromatin track.
Practical Interpretation and Common Pitfalls
Before interpreting a track, verify that the candidate and track refer to the same reference assembly and compatible coordinates. Check the assay, sample, signal definition, normalization, and peak-calling procedure. A peak absent under one threshold is not proof of absent biological activity; a weak or missing signal may reflect assay sensitivity or processing. Likewise, a strong signal is meaningful only in the context of that track’s measurement and background.
Common mistake: “The variant overlaps a strong H3K27ac peak, so it activates an enhancer and causes the association.” This reasoning turns a regional mark into a variant-specific mechanism and then into a causal claim. The correction is to state only what was observed: the candidate lies in an H3K27ac-enriched interval in the assayed material, a pattern consistent with active regulatory chromatin. Establishing a change caused by that variant, and connecting it to a trait, requires additional evidence.
A useful report makes the scope of the evidence explicit. For example: “The candidate coordinate overlaps an ATAC-seq peak and an H3K27ac-enriched interval in the specified sample, supporting regulatory activity at the region. These tracks do not establish that the candidate variant itself alters chromatin or causes the trait association.” This wording is specific, informative, and calibrated to what the assays measured.
Check Your Understanding
Answer each question by distinguishing the measured signal from the interpretation it supports.
- What does an ATAC-seq peak report, and why is it not a direct measurement of a candidate variant’s effect?
- A candidate overlaps an H3K4me1-enriched interval but has little H3K27ac signal. Give a cautious interpretation of this pattern.
- Why should normalized signal values from two different histone-mark tracks not automatically be compared directly?
- A candidate lies outside every called peak in one DNase-seq analysis. What conclusion is supported, and what stronger conclusion is not?
- State one additional kind of evidence that could address whether the candidate variant itself has a functional effect, rather than merely lying in a regulatory region.