Why Tissue Context Changes the Meaning of an Annotation
A regulatory element can be active in one cell type and inactive in another. As a result, finding a candidate variant in an enhancer annotation is not enough to interpret its relevance: the annotation’s tissue or cell type matters. An enhancer active in a plausible disease-relevant cell type may support a regulatory hypothesis, while the same genomic overlap in an unrelated tissue may be less informative. Neither observation, by itself, establishes that the variant changes regulation or causes the phenotype.
Tissue matching is a question of biological context, not a simple label-matching exercise. A phenotype may involve several organs, cell types, or developmental stages. The tissue where a phenotype is observed may not be the tissue where a causal molecular effect begins: for example, a circulating signal can affect a distant organ, and a developmental effect may not be visible in adult tissue. Matching therefore requires a reasoned link between the phenotype and the sampled biological context.
The basic inputs are a set of candidate variants at a locus, their fine-mapping results, annotations of regulatory activity, and information about the tissue, cell type, assay, and sample used to produce each annotation. A posterior inclusion probability (PIP) describes the probability assigned by a fine-mapping model that a variant is causal, conditional on that model and its assumptions. It is not a probability that a particular tissue or mechanism is responsible.
Before interpreting a tissue match, check that the variant and annotation refer to compatible genomic coordinates and alleles, and identify what the assay actually measures. Accessibility, histone marks, and regulatory-element calls are related but distinct evidence. A peak in one assay should not be paraphrased as proof of enhancer activity, gene regulation, or a phenotype effect unless those claims are independently supported.
Reasoning From Phenotype to Tissue
A useful evaluation begins by stating why a tissue or cell type is plausible for the phenotype. This rationale may come from the known physiology of the trait, the cells that produce or respond to a relevant molecule, or a well-defined disease process. It should not be inferred solely from the fact that a candidate annotation happens to be present there. That would make the annotation both the evidence and the justification for its own relevance.
The distinction is especially important for broad phenotypes. A blood measurement, for example, may reflect production in one tissue, clearance in another, and regulation by systemic signals. Conversely, a phenotype named for an organ does not guarantee that the causal effect acts only in that organ. A defensible interpretation describes the biological rationale and its limits rather than declaring one tissue “the” relevant tissue prematurely.
Worked Example: A Regulatory Annotation in a Plausible Cell Type
Suppose a hypothetical fine-mapping analysis of an airway-related phenotype considers four variants. For this example, the model assumes one causal variant among the four, so their PIPs sum to one. An annotation resource reports regulatory activity as follows.
| Variant | PIP | Hypothetical regulatory annotation |
|---|---|---|
| v1 | 0.55 | Enhancer annotation in airway epithelial cells |
| v2 | 0.25 | Accessible region in airway epithelial cells |
| v3 | 0.12 | Enhancer annotation in liver cells |
| v4 | 0.08 | No overlap in the listed tracks |
First check the stated PIPs: \(0.55+0.25+0.12+0.08=1.00\). The leading variant, v1, has PIP \(0.55\), and it overlaps an enhancer annotation in a cell type that is plausibly relevant to an airway phenotype. This is a reason to prioritize the airway regulatory hypothesis for further evaluation. It does not mean there is a 55% probability that the enhancer mechanism is correct: the PIP concerns variant causality under the fine-mapping model, not the tissue or mechanism.
The v2 accessibility annotation is also in a plausible cell type, but accessibility alone does not establish enhancer function. The liver annotation for v3 is not automatically irrelevant: a systemic or developmental pathway could make it pertinent, but a specific rationale would be needed. The absence of overlap for v4 means only that the listed tracks do not annotate it; it does not show that v4 has no regulatory function.
The appropriate conclusion is limited: the annotations make an airway epithelial regulatory explanation plausible, particularly around v1, but they do not establish a causal regulatory effect or exclude other tissues and mechanisms.
Cell-Type Specificity and Mixed Samples
Many reference datasets are collected from bulk tissue rather than purified cell types. A bulk sample combines signals from the cell populations it contains. An annotation observed in bulk tissue therefore may arise from a subset of cells, and a weak bulk signal can conceal strong activity in a rare population. Conversely, calling a bulk tissue “relevant” does not establish that the specific cell type implicated by a hypothesis is responsible for the signal.
A simple mixture calculation makes the dilution issue clear. Suppose a measured signal is on a scale where the bulk value is approximated by a cell-fraction-weighted average. Let \(f\) be the fraction of a target cell type, \(S_T\) its signal, and \(S_O\) the average signal in other cells. The idealized bulk signal is
This is a simplifying model, not a universal description of every assay. Differences in cell size, assay capture, normalization, and sample handling can make the actual relationship depart from a simple weighted average. Its key implication still matters: a bulk measurement can blend distinct cell-type signals.
Worked Example: Strong Activity Diluted in a Bulk Sample
Consider a hypothetical assay in which a target cell type makes up 20% of a tissue sample. Its regulatory signal is 0.80 on the assay’s normalized scale, while the average signal in the remaining cells is 0.10. Under the idealized mixture model,
The bulk value is 0.24, substantially below the target-cell signal of 0.80. If the target cell type instead made up 50% of the sample, keeping the cell-specific signals unchanged, the bulk value would be
Thus, a modest bulk signal does not rule out strong activity in a rare cell type. Nor does the calculation establish that the target cells actually carry the signal at a locus: that requires relevant assay data. The calculation only shows why cell composition must be considered when interpreting bulk measurements.
Bulk tissue also creates an attribution problem. If a peak is present in a mixed sample, the assay may not identify which constituent cells produced it. Cell composition can vary across donors, disease states, and experimental preparations, further complicating comparisons. A tissue label should therefore be read at the resolution the data support: “bulk lung” is not equivalent to “airway epithelial cells,” and “whole blood” is not equivalent to a particular immune-cell subtype.
Incomplete Reference Panels
Reference panels rarely contain every potentially relevant tissue, cell type, developmental stage, or activation state. A missing annotation can therefore have two different meanings: the feature was measured and not detected, or it was not measured in that context. These are not interchangeable. A non-detection in a well-matched assay is evidence that the feature was not observed under those conditions; an unmeasured tissue provides no direct evidence of absence there.
Even within a measured tissue, a negative result depends on assay sensitivity, sample composition, biological state, and the way the annotation was defined. Absence of a peak in a particular dataset is not proof that the element is never active. Positive and negative results should both be described with their experimental context.
Worked Example: The Relevant Cell Type Is Not in the Panel
Suppose a hypothetical locus is associated with a phenotype of airway mucus production. The available regulatory panel contains tracks for liver, whole blood, and skin fibroblasts, but no airway epithelial sample. A candidate variant overlaps an annotated enhancer in liver and has no overlap in the other two tracks.
The panel contains 3 tissue contexts, but 0 of those 3 is the hypothesized airway epithelial context. The observed liver overlap is a real annotation in the panel, but its relevance to the phenotype requires an independent biological rationale. The two non-overlaps do not test whether the candidate variant lies in an active element in airway epithelial cells, because that context was not assayed.
The correct conclusion is that the panel provides a liver enhancer annotation and no overlap in the measured blood and fibroblast tracks; it does not resolve airway epithelial activity. It would be incorrect to say that the variant has no relevant regulatory annotation simply because the panel lacks the most plausible cell type. The result is a limitation of coverage, not a negative test of the unmeasured context.
A Practical Evaluation Sequence
The following sequence keeps biological relevance separate from what the assays directly establish.
Describe the relevant physiology or disease process and why a tissue or cell type might participate. Keep plausible alternatives open when the phenotype could involve more than one context.
Record whether the annotation comes from a purified cell type or a mixed tissue, and distinguish the assay’s direct measurement from its interpretation as a regulatory element.
Check coordinate compatibility and determine which candidate variants overlap which annotations. Interpret fine-mapping results as variant evidence, not as a tissue-specific probability.
Ask whether a signal could be diluted or attributed to an unknown cell population, and whether plausible contexts were measured at all.
State which tissue-specific hypothesis is supported, what the data directly show, and what remains uncertain. Do not turn a plausible match into proof of mechanism.
Why Tissue Matching Matters—and Where It Can Mislead
Tissue context helps prioritize functional hypotheses because regulatory activity is often conditional on cell type and biological state. It can distinguish an annotation that is biologically plausible for a phenotype from one whose relevance is unclear, and it can point to the kinds of samples needed for follow-up. Its value is in sharpening interpretation, not making the evidence stronger than the experiment warrants.
A strong tissue match is still not a complete causal chain. It does not, on its own, show that the candidate variant changes the annotation, that the affected element regulates a particular gene, or that the molecular change produces the phenotype. Those are separate links that require their own evidence. Conversely, a weak match may reflect incomplete panels or mixed samples rather than a biologically irrelevant locus.
Common Mistake
- Wrong reasoning: “The candidate variant overlaps an enhancer in the tissue named by the phenotype, so that tissue and enhancer must mediate the association.”
- Correction: The overlap is supportive context, not proof. Check what the assay measured, whether the sample resolves the relevant cell type, whether the annotation is biologically plausible for the phenotype, and whether other explanations remain possible.
- Related mistake: “There is no overlap in the available tracks, so the variant is not regulatory.” If relevant tissues or states were not assayed, that conclusion goes beyond the data. Report the missing coverage explicitly.
Check Your Understanding
Answer each question by distinguishing what the annotation shows from what it does not establish.
- A variant overlaps an accessible region in a cell type plausibly involved in a phenotype. What is supported, and what additional claims would be unjustified from this overlap alone?
- Why might a strong cell-type-specific regulatory signal produce only a modest signal in a bulk tissue assay?
- A candidate variant has no annotation in a reference panel that did not assay the most plausible cell type. What conclusion can be drawn about that cell type?
- Why is “whole blood” not necessarily an adequate substitute for a named immune-cell subtype when evaluating a locus?
- How should a PIP be distinguished from the probability that a particular tissue or regulatory mechanism explains the phenotype?