Why Probe Design Can Change a Genotype Signal
A genotyping array infers alleles from binding signals at designed probes. The earlier material on reading genotype clusters from array intensities establishes how these signals can form patterns associated with the three diploid genotypes. Those patterns depend, however, on the probes binding the intended DNA sequence in a sufficiently consistent way. A nearby sequence difference or binding to an unintended sequence can change the pattern even when the genotype at the target site has not changed.
The target variant is the position whose genotype the array is intended to measure. A probe-design error in this tutorial means a failure of the probe’s sequence or binding behavior to represent that target reliably. The term covers several mechanisms: an unaccounted-for variant near the target can weaken probe binding; a poor match between probe and intended target can reduce signal; and a probe can bind similar DNA elsewhere in the genome. These mechanisms can lower a signal, shift the apparent balance between allele signals, or add signal where the target allele is absent.
The effect is not always random. If a nearby variant is carried on the same chromosome as one target allele, it can selectively affect the probe signal associated with that allele. As a result, an artifact may recur among samples carrying the same local sequence, rather than appearing as unrelated noise. The observed signal alone does not establish whether an unusual genotype is biological or technical.
A Simple Model of Allele-Specific Intensities
Let \(I_A\) and \(I_B\) denote the measured signals from probes associated with target alleles A and B. As a useful simplification, suppose background has been subtracted, there is no binding to unintended DNA, and the signal from each allele is proportional to the number of corresponding target copies. Write \(n_A\) and \(n_B\) for those copy counts, and \(e_A\) and \(e_B\) for the respective probe efficiencies. Then
In a diploid sample, the intended copy counts are \((n_A,n_B)=(2,0)\) for AA, \((1,1)\) for AB, and \((0,2)\) for BB. Under equal efficiencies, a heterozygote has balanced signals in this simplified model. In practice, signals also include background and measurement variation, and platforms use their own intensity processing and calling procedures. The model is a way to reason about direction of effect, not a universal calibration rule.
An illustrative measure of signal balance is the contrast
when the total signal \(I_A+I_B\) is positive. A positive \(M\) means the A-associated signal is larger; a negative \(M\) means the B-associated signal is larger. The contrast does not, by itself, determine a genotype: total intensity and the platform’s calling model also matter.
How a Mismatch Shifts the Contrast
For an AB sample with no background or cross-hybridization, the simplified model gives \(I_A=e_A\) and \(I_B=e_B\). Substitution into the contrast gives
Thus, even though both alleles are present, unequal efficiencies move the contrast away from zero. If a nearby variant lowers \(e_A\), the A-associated signal becomes relatively weaker. If it lowers both efficiencies, total intensity can fall as well. Depending on the calling procedure and the pattern in the data, these changes can make a sample harder to classify or place it near a different genotype cluster.
Worked Example: A Nearby Variant Weakens One Allele Signal
Suppose a hypothetical AB sample has background-adjusted intensities \(I_A=102\) and \(I_B=98\) when both probes match the sample sequence well. Its contrast is
The two signals are similar, as expected for this illustrative heterozygote. Now suppose a nearby variant on the chromosome carrying allele A reduces the A-probe intensity to 45, while the B-probe intensity remains 98. The observed contrast becomes
The contrast has moved from slightly positive to negative, even though the target genotype remains AB. The combined intensity has also fallen from \(200\) to \(143\) arbitrary intensity units. This combination of a changed balance and lower total signal is consistent with an allele-specific binding problem, but it does not prove one: other sources of signal variation could produce similar observations. A calling algorithm might classify this sample as uncertain, assign a misleading call, or make a no-call, depending on its model and thresholds.
Cross-Hybridization: Signal From the Wrong Sequence
A probe may bind DNA sufficiently similar to its intended target even when that DNA comes from another genomic location. This is called cross-hybridization. Similar sequence, including sequence in related or repeated regions, can contribute signal to one or both allele channels. The consequence depends on which probe binds the unintended sequence and how strongly it binds.
Cross-hybridization is different from a nearby variant weakening a probe. A mismatch can reduce signal from the intended target; cross-hybridization can add signal from elsewhere. Either can distort allele balance. If unintended binding contributes to a channel associated with an allele absent at the target, the signal may resemble evidence for that allele. If it contributes similarly to both channels, it may increase total intensity without creating a clear allele-specific shift.
Worked Example: Off-Target Binding Creates an Apparent Second Allele
Consider a hypothetical AA sample. After background adjustment, suppose its intended target contributes 100 units to the A-associated signal and 5 units to the B-associated signal. The small B value represents residual signal, not a B copy at the target. Without other binding, the contrast is
Now suppose an unintended, similar sequence contributes 70 units to the B-associated channel and no additional A signal. The measured signals are then \(I_A=100\) and \(I_B=5+70=75\), giving
The B-associated channel is now substantial, so the signal could look more like evidence for a second allele than the original measurement did. The example does not imply that this contrast must be called AB: total intensity, the array’s cluster pattern, and its calling procedure matter. It shows why a signal in an allele-associated channel is not conclusive evidence that the corresponding allele is present at the target site.
Worked Diagnosis and Practical Checks
Diagnosing probe-related errors means comparing the observed pattern with the probe’s intended target and with other available evidence. A single unusual intensity is rarely enough to identify the mechanism. Check whether the anomaly is mainly low total signal, a shifted allele balance, or both; then investigate whether the probe sequence and binding context could explain it.
Worked Example: Choosing Checks for a Low-Signal Cluster
In a hypothetical array run, most samples at one marker form three familiar clusters. Four samples have much lower total intensity than the rest and sit between the expected clusters. The four samples also share the same sequence variant near the probe-binding region, as established by an independent assay.
First, describe the evidence. The low total intensity and intermediate positions are consistent with difficulty assigning a clear genotype. Their shared nearby variant provides a plausible mechanism: it may reduce binding to one or both allele-specific probes. This is a hypothesis supported by the pattern, not a conclusion from intensity alone.
Next, compare relevant signals. If one allele-associated signal is selectively reduced in samples carrying the nearby variant, that supports an allele-specific effect. If both signals fall, the variant may affect both probe interactions, or some other sample-level factor may be involved.
Then, test the interpretation. Review the probe sequence and the position of the nearby variant, check replicate samples if available, and compare with an independent genotyping method. A probe targeting a different local sequence can also be informative. If the independent result agrees with the intended genotype while this array marker remains weak, the array signal is a likely source of the ambiguity. If results disagree, further investigation is needed; neither result should be declared correct solely because it is more convenient.
Inspect total signal and allele balance, and note whether affected samples are scattered or share a consistent pattern.
Check the intended probe sequence for nearby variation and consider whether similar sequences elsewhere could bind the probe.
Use replicates, an alternative probe, or an independent genotyping assay when available; agreement and disagreement each need contextual interpretation.
Common Mistakes, Alternatives, and Limitations
- Mistake: treating an allele-channel signal as proof of that target allele. Cross-hybridization can contribute signal from another sequence. Check probe specificity and, where consequential, seek independent confirmation.
- Mistake: assuming every mismatch causes the same effect. The effect depends on the probe design and sequence context; a nearby difference may weaken one signal, both signals, or have little measurable effect. Examine the observed pattern rather than applying a universal rule.
- Mistake: interpreting a no-call as evidence for a particular genotype. A no-call indicates that the calling procedure did not assign a genotype under its criteria. It does not identify why the signal was ambiguous.
- Mistake: responding to a shifted cluster only by relaxing call thresholds. A more permissive threshold can produce more calls without correcting the underlying signal distortion. It may turn uncertain measurements into misleading assignments.
Possible responses serve different purposes. Replacing or redesigning a probe can address a problematic binding sequence, but the new design must itself be evaluated. Masking an unreliable marker or assigning no-calls can avoid using uncertain calls, at the cost of losing data at that marker or in affected samples. Independent genotyping can help resolve important disagreements, though it adds work and does not remove the need to assess the reliability of each method. These choices are not interchangeable: improving a probe, avoiding an unreliable call, and checking a genotype answer different questions.
Intensity diagnostics have limits. A cluster plot can reveal unusual signal structure, but it usually cannot identify a nearby variant or prove cross-hybridization on its own. A genuine rare genotype can also lie away from common clusters, and a technical artifact can form a consistent-looking pattern. Probe sequence review, sample context, and independent evidence are therefore complementary. The central lesson is to treat genotype calls as measurements whose reliability depends partly on how well the probe represents the intended target.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions.
- An AB sample has background-adjusted signals \(I_A=80\) and \(I_B=120\). Calculate \(M\). What does its sign say, and what does it not establish?
- A nearby variant reduces the efficiency of the A-associated probe in an AB sample. Under the simplified model, which signal and which direction of contrast change are expected?
- How can cross-hybridization affect an AA sample differently from a mismatch that only reduces the A-probe signal?
- Why is it not sufficient to lower a genotype-calling threshold when many samples at one marker have weak, shifted signals?
Solutions. (1) \(M=(80-120)/(80+120)=-40/200=-0.200\). The B-associated signal is larger, but the sign alone neither establishes a BB genotype nor identifies the reason for the imbalance. (2) The A-associated signal falls relative to the B-associated signal, so \(M\) moves in the negative direction; total intensity may also fall if the B signal is unchanged. (3) Cross-hybridization can add signal from an unintended sequence, potentially increasing an allele-associated channel even though that allele is absent at the target. A mismatch that selectively weakens the A probe instead reduces the intended A-associated signal. (4) A lower threshold may assign calls to distorted or ambiguous measurements without fixing their cause. Probe context and independent evidence should be assessed, and uncertain calls may need to remain unassigned.