Promoters, Enhancers, and Regulatory-Element Overlap

Why Regulatory-Element Overlap Needs Care

A fine-mapped variant can fall within a promoter or enhancer annotation, giving a plausible route by which it might affect gene regulation. But an overlap is first a coordinate statement: the variant’s genomic span and the annotated interval share one or more bases. It is not, by itself, evidence that the variant changes regulatory activity, that the element regulates a particular gene, or that the variant is causal.

Making that coordinate statement correctly requires knowing how the annotation was defined and ensuring that variant and element coordinates use the same genome assembly and coordinate convention. A boundary error of one base can change a reported overlap, especially for a single-nucleotide variant (SNV). The interpretation also depends on whether an element label describes a narrow experimentally supported interval or a broader region assigned by an annotation pipeline.

What Promoter and Enhancer Annotations Represent

Definition: A promoter annotation is a genomic interval designated as promoter sequence, usually by its position relative to a transcript’s transcription start site (TSS) and sometimes by supporting assay evidence. An enhancer annotation is a genomic interval designated as enhancer sequence by an assay, an annotation pipeline, or both. These are operational labels: their exact boundaries and evidence requirements depend on the source and method.

A promoter is commonly defined using a window around a TSS, but the size and placement of that window are not universal. A promoter annotation might include bases upstream of the TSS, downstream bases, or both. “Upstream” is relative to the direction of transcription: for a transcript on the reverse strand, upstream is toward increasing genomic coordinates. Therefore, a promoter interval should be interpreted using its recorded coordinates and transcript context, not by assuming that it always lies at lower coordinate values than the gene.

Enhancers are regulatory elements that can influence transcription, often from a distance, but a database label does not establish which gene an element regulates. An enhancer interval may come from an assay designed to identify regulatory activity, a pipeline that combines assay evidence, or a standardized annotation resource. The interval is the method’s reported region; it is not necessarily a precise map of every functional base. Different methods can give different boundaries or labels for the same locus.

For a defensible overlap analysis, record at least the genome assembly, the source of the annotation, the interval coordinates, and the transcript or assay context used to define it. A coordinate match is only meaningful if the variant and annotation have been mapped to the same assembly. The earlier material on variant coordinates and allele harmonization establishes why variant identity, reference alleles, and genomic coordinates must be checked before annotation.

Coordinate Conventions and the Overlap Rule

Many variant tables report one-based positions, while BED files conventionally use zero-based, half-open intervals. A one-based SNV at position $p$ corresponds to the BED interval $[p-1,p)$. A BED interval $[a,b)$ includes its start coordinate and excludes its end coordinate; it represents $b-a$ bases. In one-based base positions, that BED interval covers positions $a+1$ through $b$, inclusive.

Definition: Two BED intervals $[a,b)$ and $[c,d)$ overlap if their intersection has positive length. The intersection length is $$ L=\max\left(0,\min(b,d)-\max(a,c)\right). $$ If $L=0$, the intervals share no bases. If $L>0$, they share $L$ bases.

The formula follows from the endpoints of the intersection. Its left endpoint is the later of the two starts, $\max(a,c)$, and its right endpoint is the earlier of the two ends, $\min(b,d)$. Subtracting the left endpoint from the right gives the intersection length when the right endpoint is larger. If it is equal or smaller, there is no positive-length intersection, so the maximum with zero gives $L=0$. In particular, intervals that meet exactly at an excluded end coordinate have zero overlap.

For an SNV, the BED interval has length one, so an overlap is either zero bases or one base. A multi-base variant can overlap only part of an annotated element, or only part of its own reference span. For a variant represented by a reference allele of length $k$ beginning at one-based position $p$, the reference span is commonly represented as the BED interval $[p-1,p-1+k)$. This describes the span used for positional annotation; it does not mean every base in that span is functionally altered in the same way.

Worked Example: An SNV at the BED End Coordinate

Suppose a hypothetical enhancer annotation is BED interval $[499,520)$, and a fine-mapped SNV is reported at one-based position 520. Assume both coordinates refer to the same genome assembly and the SNV position has been verified.

Convert the SNV to BED coordinates: position 520 becomes $[519,520)$. The enhancer interval includes BED coordinates 499 through just before 520, corresponding to one-based positions 500 through 520. The overlap length is

L=\max\left(0,\min(520,520)-\max(499,519)\right) =\max(0,520-519)=1.

Thus, the SNV overlaps the enhancer by one base. The fact that the BED end is 520 does not exclude one-based position 520: BED’s end coordinate is exclusive, but BED base 519 corresponds to one-based position 520. The result is a positional match, not evidence that the SNV alters enhancer activity.

Worked Example: A Variant at and Beyond a Promoter Boundary

Consider a hypothetical promoter interval $[2000,2015)$ in BED coordinates. For this example, treat the interval as the promoter annotation supplied by the source; assume its assembly matches the variant data. Compare SNVs at one-based positions 2015 and 2016.

The interval $[2000,2015)$ contains one-based positions 2001 through 2015. The SNV at position 2015 maps to $[2014,2015)$, giving

L_{2015}=\max\left(0,\min(2015,2015)-\max(2000,2014)\right)=1.

It overlaps the promoter by one base. The SNV at position 2016 maps to $[2015,2016)$, so

L_{2016}=\max\left(0,\min(2015,2016)-\max(2000,2015)\right)=0.

This second SNV starts exactly at the promoter’s exclusive end coordinate, so the intervals share zero bases. The promoter annotation itself does not imply that the variant at 2016 is biologically inactive; it means only that this variant does not overlap this particular interval under the stated coordinates.

Partial Matches and Multiple Labels

For an SNV, overlap with a single interval cannot be fractional: the SNV contributes either one shared base or none. Partial overlap becomes possible for a multi-base variant span, an interval that covers only part of that span, or a broad element whose boundary intersects a larger variant. It is useful to report both the shared length and, where informative, the fraction of the variant span or element that is shared. The denominator must be stated: “one base overlaps” is not the same statement as “one third of the variant span overlaps.”

Worked Example: Partial Overlap of a Multi-Base Variant

Suppose a hypothetical three-base variant has a reference span at one-based positions 300–302, inclusive. Its BED span is $[299,302)$. An enhancer annotation is $[300,301)$, a one-base interval corresponding to one-based position 301. Assume the variant and annotation use the same assembly and that this interval is the annotation being evaluated.

The intersection begins at $\max(299,300)=300$ and ends at $\min(302,301)=301$. Therefore,

L=\max(0,301-300)=1 \text{ base}.

One of the three bases in the variant’s reference span overlaps the enhancer, so the overlap covers $1/3$ of that span. This fraction describes coordinate overlap only. It does not establish that the overlapping base is the only functionally relevant base, nor that the variant changes enhancer function.

A variant can also overlap more than one annotated element. Promoter and enhancer annotations need not be disjoint: sources can use different definitions, and distinct annotations can overlap in genomic coordinates. A single variant may therefore receive multiple labels without any coordinate contradiction. Preserve each overlap and its source rather than forcing the variant into one mutually exclusive category.

Worked Example: One SNV Overlaps Two Annotations

Suppose a hypothetical SNV at one-based position 805 maps to BED interval $[804,805)$. A promoter annotation is $[800,806)$, and an enhancer annotation is $[804,810)$. Assume all intervals use the same genome assembly.

For the promoter, the intersection is $[804,805)$, so the overlap length is $\min(806,805)-\max(800,804)=1$ base. For the enhancer, the intersection is also $[804,805)$, giving $\min(810,805)-\max(804,804)=1$ base. The SNV overlaps both annotations by one base.

A clear report retains both labels and identifies their annotation sources. It should not conclude that the variant is simultaneously proven to act as a promoter and an enhancer, or that either annotation identifies its target gene. The overlap records the result of comparing the SNV span with two reported intervals.

Connecting Overlaps to Fine-Mapping

Fine-mapping assigns candidate variants posterior inclusion probabilities (PIPs), which quantify the model-based probability that each variant is causal under the fine-mapping model and its assumptions. An overlap annotation adds a different kind of information: it describes the location of a candidate relative to an annotated interval. Keeping these quantities distinct avoids turning a positional match into a causal claim.

For example, if a variant with a PIP of 0.20 overlaps an enhancer annotation, the PIP remains 0.20; the overlap does not automatically increase it. The annotation may be biologically relevant evidence for prioritizing follow-up, but changing the fine-mapping probability requires a model that incorporates such evidence. Likewise, a variant outside a particular annotation is not thereby shown to be nonfunctional: the annotation may be incomplete, or the relevant element may not have been included in that source.

When describing results, identify the variant, element class, annotation source, coordinate convention, and overlap length. If multiple candidate variants overlap, do not treat the number of overlaps alone as a probability that the element is causal. Variants in a credible set can have different PIPs, and the meaning of a set-level summary requires care beyond simply counting matches.

Common Mistakes and Checks

Mixing coordinate systems. Comparing a one-based position directly with BED start and end coordinates can create an off-by-one error. Convert both to a common convention before calculating overlap.
Treating the BED end as included. In $[a,b)$, coordinate $b$ is excluded. A variant beginning at $b$ has no shared bases with that interval.
Calling a boundary match “no overlap” without calculating it. A one-based SNV at position 520 maps to $[519,520)$; it overlaps $[499,520)$ by one base. Calculate the intersection rather than relying on a visual comparison of endpoint labels.
Assuming all promoter or enhancer definitions are universal. Promoter windows and enhancer boundaries depend on the source and method. Report which annotation was used.
Interpreting overlap as causality or a gene link. Overlap is positional evidence. It does not by itself show a molecular effect, causal status, or target gene.
Discarding multiple labels as inconsistent. Distinct promoter and enhancer annotations can overlap. Retain the labels and their provenance rather than imposing mutually exclusive categories.

A reliable analysis checks assembly and coordinate convention first, verifies how each element interval was defined, calculates intersection length explicitly, and then states only what that calculation supports. These checks are particularly important for SNVs at element boundaries, where a single base determines whether the result is an overlap.

Key takeaway: Promoter and enhancer overlap is an annotation-dependent coordinate relationship. Convert variant and element intervals consistently, calculate shared bases with half-open interval rules, and interpret the match as functional context—not as proof of regulatory effect, causality, or a target gene.

Check Your Understanding

Use a common coordinate convention when answering, and distinguish an overlap result from a functional conclusion.

A BED enhancer is $[99,105)$. Does an SNV at one-based position 105 overlap it? Convert the SNV and calculate the shared length.
A BED interval is $[40,44)$. Which one-based positions does it cover, and how many bases does it contain?
A variant spans three reference bases, but only one base lies within an annotated promoter. What is the overlap length, and what fraction of the variant span overlaps?
Why can a single variant receive both promoter and enhancer overlap labels without the coordinate analysis being contradictory?
A fine-mapped variant overlaps an enhancer annotation. What does that establish, and what does it not establish about causality or the target gene?