Immunoassays interview questions

What a strong answer covers

• In a direct ELISA the antigen is immobilised and detected with a labelled primary antibody. It is quick and has few steps, but sensitivity is limited and every primary must be conjugated.
• An indirect ELISA adds a labelled secondary antibody against the primary, which amplifies signal and lets one labelled secondary serve many primaries. It is the standard format for measuring antibody titres in serum, and it introduces the possibility of secondary antibody cross-reactivity.
• A sandwich ELISA captures the analyte between two antibodies recognising different epitopes. It is the most specific and sensitive format for measuring a protein analyte in a complex matrix, and it requires the analyte to be large enough to present two non-overlapping epitopes at once.
• A competitive ELISA measures signal that decreases as analyte increases, because sample analyte competes with a labelled or immobilised analogue. It is the format of choice for small molecules and haptens that cannot support a sandwich, and for analytes with only one usable epitope.
• The selection logic is therefore driven by the analyte: size and epitope availability decide whether a sandwich is possible, expected concentration decides how much sensitivity is needed, and the matrix decides how much specificity is needed.
• A strong answer notes that the same logic carries over to other platforms such as bead-based multiplex and electrochemiluminescence assays, where the formats are the same and the detection chemistry differs.

Where people lose the point

• ×Proposing a sandwich assay for a small hapten that cannot bind two antibodies simultaneously.
• ×Treating the choice as a matter of habit or kit availability rather than of analyte properties.
• ×Forgetting that in a competitive format the signal is inversely related to concentration, which inverts the interpretation of the curve.

What a strong answer covers

• Blocking occupies the remaining protein-binding sites on the plate surface after coating, so that antibodies added later bind their targets rather than the plastic. Without it, background rises across the plate and the assay loses its ability to distinguish low positives.
• The blocking agent is a real variable, not a formality: protein blockers and detergent-based blockers behave differently, and a blocker that shares an epitope with the detection system (for example a protein blocker in an assay targeting that protein) will cause trouble.
• Washing removes unbound reagent between steps. Insufficient washing leaves detection reagent behind and raises background, while over-aggressive washing or allowing wells to dry can strip bound material and lower signal.
• Consistency across the plate matters as much as thoroughness. Uneven washing is a classic cause of positional patterns in results, particularly at plate edges.
• A strong answer mentions soak time and number of cycles as parameters that should be defined in the method, and that a plate washer needs maintenance because blocked manifold channels produce reproducible, misleading patterns.

Where people lose the point

• ×Treating wash steps as unimportant technique detail rather than a defined, validated parameter.
• ×Letting wells dry out between steps.
• ×Using a blocking agent that interacts with the analyte or the detection antibody.

What a strong answer covers

• The response of a sandwich immunoassay is sigmoidal against log concentration, bounded by a background floor and a saturation ceiling, so a four-parameter logistic fit is the usual model and a five-parameter fit is used where the curve is asymmetric.
• Evaluate the fit by back-calculating the standards through it. If a standard does not recover close to its nominal value, the fit does not describe the assay at that point, regardless of how good the correlation coefficient looks.
• The quantitative range is bounded by where accuracy and precision both remain acceptable. The lower boundary is the lowest concentration measured with defined accuracy and precision, which is distinct from the limit of detection, the lowest concentration distinguishable from blank.
• Samples reading above the top of the range are diluted and re-run so that the reading falls in range, and the result is multiplied by the dilution factor. Extrapolating beyond the highest standard is not a legitimate substitute.
• Standards should be prepared in a matrix that resembles the samples wherever possible, because a curve built in buffer and applied to serum imports a matrix difference into every result.
• A strong answer discusses replicate variability and acceptance criteria, and mentions weighting the fit so that low-concentration points are not dominated by the high end.

Where people lose the point

• ×Judging a curve by its correlation coefficient rather than by back-calculated recovery of the standards.
• ×Reporting values extrapolated beyond the top or bottom standard.
• ×Confusing the limit of detection with the lower limit of quantitation.

What a strong answer covers

• A matrix effect is any influence of the sample's composition, other than the analyte, on the measured result. Serum, plasma, cell culture supernatant, urine and tissue lysate each present different protein content, viscosity, salt, lipid and additive load.
• Spike and recovery tests it directly: add a known amount of analyte to the real matrix and to the assay diluent, and compare recovered against expected. Poor recovery in matrix and good recovery in diluent is a matrix effect.
• Dilutional linearity, or parallelism, tests whether a high sample diluted through the range gives concentrations that agree after correcting for dilution. Non-parallel behaviour indicates the matrix or the analyte form is interacting with the assay.
• Common mechanisms include binding proteins that sequester the analyte, anticoagulant effects that differ between serum and plasma tube types, endogenous antibodies that bridge or block the assay antibodies, and lipaemia or haemolysis interfering with detection.
• Mitigations include a matrix-matched calibrator, a minimum required dilution that dilutes out the interference, a sample pretreatment step, or adding blocking reagents that neutralise interfering endogenous antibodies.
• A strong answer notes that sample type should be fixed and validated: a method validated in serum does not automatically transfer to plasma or to a different anticoagulant.

Where people lose the point

• ×Validating in buffer only and assuming it transfers to biological matrix.
• ×Treating spike and recovery and dilutional linearity as the same experiment when they answer different questions.
• ×Mixing serum and plasma samples in one study without demonstrating equivalence.

What a strong answer covers

• In a one-step sandwich format, sample and detection antibody are present together. At extremely high analyte concentration, the analyte saturates the capture antibody and the detection antibody independently, so the bridging complex does not form in proportion and signal falls instead of rising.
• The result is a dose response curve that rises, peaks and then declines, so an extremely high sample can return a value that appears normal, with nothing on the plate to indicate a problem.
• The most direct detection is dilution: a sample affected by the hook effect gives a higher calculated concentration on further dilution, which is the reverse of the expected behaviour and is diagnostic.
• Design defences include a two-step format with a wash between sample incubation and detection antibody addition, an antigen excess check built into the assay, or a defined policy of routinely running clinically plausible high samples at more than one dilution.
• A strong answer connects it to the human factor: the risk is highest when a result that looks unremarkable contradicts a strong clinical or experimental expectation, and the correct response is to re-run diluted rather than to accept the number.

Where people lose the point

• ×Describing it as a saturation plateau rather than a decline in signal at very high concentration.
• ×Assuming every assay format is susceptible, when a two-step wash format largely removes the mechanism.
• ×Trusting an in-range result on a sample where everything else says the analyte should be extreme.

What a strong answer covers

• The pair must bind non-overlapping epitopes so both can engage the analyte at once, which is why monoclonal pairs are usually screened in a matrix of combinations rather than reasoned about from datasheets alone.
• Screen both orientations. A pair often performs very differently depending on which antibody is coated and which is the detector, because immobilisation can occlude a binding site.
• Affinity matters in specific ways: a high-affinity capture antibody supports sensitivity, while the detection antibody's affinity and labelling ratio influence signal and background. Over-labelling a detection antibody can raise background more than signal.
• Specificity screening should include the obvious confounders early: related family members, precursor and degradation forms, and the species used in the model system if the assay must be species specific.
• Decide which form of the analyte the assay should measure. Free versus bound, total versus active, and cleaved versus intact are genuine design decisions, and the pair's epitopes determine which one you get.
• A strong answer includes practical screening outputs: signal to background at relevant concentrations, the shape of the resulting curve, and performance in real matrix rather than buffer alone.

Where people lose the point

• ×Selecting a pair on datasheet claims without a cross-titration screen in both orientations.
• ×Ignoring which molecular form of the analyte the epitopes select for.
• ×Optimising signal alone rather than signal relative to background in real matrix.

What a strong answer covers

• Accuracy or trueness: how close measured values are to the expected value, typically demonstrated with spiked recovery and, where available, a reference material or comparison method.
• Precision at more than one level: repeatability within a run and intermediate precision across runs, days, operators, instruments and reagent lots, since lot-to-lot variability is a common real-world failure.
• Sensitivity, distinguishing the limit of detection from the lower limit of quantitation, and defining the quantitative range where accuracy and precision both hold.
• Specificity and selectivity: cross-reactivity against related analytes and interference from common endogenous substances such as haemolysis, lipaemia, bilirubin and, for some assays, biotin or heterophilic antibodies.
• Stability: analyte stability in the sample under the storage and freeze-thaw conditions the study will actually use, and reagent and calibrator stability over their claimed shelf life.
• Robustness, meaning tolerance of small deliberate variations in incubation time, temperature and wash conditions, which is what tells you whether the method will survive a different operator on a different day.
• A strong answer states clearly that which parameters are required, and what acceptance criteria they must meet, depends on the intended use and the applicable regulatory framework, so the plan is written against that context rather than a generic template.

Where people lose the point

• ×Reciting a parameter list without connecting it to intended use or acceptance criteria.
• ×Asserting specific numeric acceptance limits as universal when they are set by the applicable guidance and the assay's purpose.
• ×Omitting lot-to-lot and operator-to-operator variation, which is where methods usually fail after transfer.

What a strong answer covers

• Start by noting that uniform high signal including blanks points at the detection end of the assay rather than at the sample end, which immediately narrows the search.
• Check the substrate and the conjugate first: substrate that has been exposed to light or contaminated, conjugate at too high a concentration, or a development step left far too long are the most common causes.
• Then check for contamination and carryover: a contaminated wash buffer, a splash of conjugate across the plate, or a reagent trough reused between steps.
• Then washing: insufficient cycles, a blocked washer manifold, or an omitted wash step leaves unbound conjugate everywhere.
• Then blocking: an inadequate or inappropriate blocker leaves the plastic available to bind detection reagent directly.
• Distinguish the pattern before concluding: uniform high signal, a gradient across the plate, edge effects, or a channel-shaped stripe each point at different causes, and the plate layout itself is evidence.
• A strong answer describes the resolution as a controlled experiment: change one variable at a time on a small plate with substrate-only and conjugate-only wells, rather than repeating the whole assay and hoping.

Where people lose the point

• ×Repeating the plate unchanged and treating a better second result as the answer.
• ×Changing several reagents at once, so the cause is never identified and the failure recurs.
• ×Ignoring the spatial pattern on the plate, which is often the most informative piece of evidence available.

What a strong answer covers

• Take it seriously and start with the sample rather than defending the assay: identity and labelling, tube type and anticoagulant, collection and storage conditions, and whether the sample was haemolysed, lipaemic or previously thawed.
• Re-run the sample, and re-run it at more than one dilution. Non-parallel results on dilution point at interference, and a higher result on greater dilution points at the hook effect.
• Consider interference explicitly: heterophilic or human anti-animal antibodies can bridge the capture and detection antibodies and generate a false positive, or block them and generate a false negative. Blocking reagents or an alternative method can test this.
• Consider whether the assay is measuring the form of the analyte that matters here. Free versus bound, or a degradation fragment recognised by one antibody, can produce a technically correct result that answers a different question.
• Check the run itself: controls, calibration, plate layout and whether the well in question sits somewhere with a known positional effect.
• A strong answer reports the outcome honestly whichever way it falls, including confirming that the assay result stands, and records the investigation so that a recurring interference pattern becomes visible over time.

Where people lose the point

• ×Defending the number because the controls passed, when controls do not carry the interference present in a patient or study sample.
• ×Amending a result without an investigation, because someone senior expected a different answer.
• ×Failing to consider that the assay may be measuring a different molecular form than the one being discussed.