ABC-BIO common mistakes
ABC-BIO · Understand the traps and how to avoid them.
When asked to choose a calibration model for a bioanalytical assay, the candidate fits a linear regression through the origin by default and reports the LLOQ from the lowest non-zero standard, without comparing weighted (1/x²) versus unweighted fits or checking back-calculated standard accuracy.
Many candidates carry over a habit from general analytical chemistry where a simple unweighted line is assumed adequate; they do not appreciate that bioanalytical calibration spans several orders of magnitude, so unweighted least squares over-weights the high end and the low standards fail accuracy criteria.
What to do instead
Before selecting a model, tabulate percent deviation of back-calculated concentrations at each standard level for unweighted and 1/x² (and 1/x) fits, then choose the model whose lowest standard meets the ±20% accuracy window while the others meet ±15%; state the weighting explicitly in the answer.
The candidate defines the lower limit of quantification as the concentration of the lowest calibration standard or as the mean of the blank plus three standard deviations, rather than as the lowest concentration validated to meet both a defined signal-to-noise threshold and accuracy/precision criteria.
The LOD formula (blank + 3 SD) is heavily drilled in general analytical chemistry and gets misapplied to bioanalysis, where LLOQ is a validated performance claim, not a statistical detection estimate.
What to do instead
Write the LLOQ definition as: lowest concentration with S/N of at least 5 (or equivalent), analyte response distinguishable from blank, and accuracy within ±20% with precision ≤20% across at least five replicates; keep LOD and LLOQ as separate concepts.
When asked to assess matrix effects, the candidate reports only the spike recovery from the post-extraction addition and concludes there is no matrix effect, omitting the comparison of post-extraction-spiked versus neat-solution responses and the variability of the internal standard response across lots.
Recovery and matrix effect are conflated because both involve spiking; candidates forget that recovery is about extraction efficiency while matrix effect is about ionization suppression/enhancement, and the internal-standard-normalized matrix factor is what reveals lot-to-lot variability.
What to do instead
Use the three-set design (neat standard, post-extraction spike, pre-extraction spike) and compute matrix factor as post-extraction response divided by neat response, plus the IS-normalized matrix factor; report both mean and variability across at least six individual lots.
The candidate evaluates carryover by injecting a single blank after the highest calibrator and calls it acceptable if any peak is present but small, without applying a defined acceptance limit relative to the LLOQ response.
Carryover acceptance is often remembered only qualitatively ('should be low'), and candidates do not recall that the blank response must typically be no more than about 20% of the LLOQ response (and IS no more than about 5%).
What to do instead
State the numeric criterion up front: blank response ≤20% of LLOQ analyte response and ≤5% of IS response, tested after the highest standard in at least duplicate, then interpret the observed data against that number.
When calculating a stability result, the candidate compares stressed and reference samples using percent difference of the means only, ignoring whether the comparison is against a freshly prepared calibration curve and whether the acceptance window is ±15% of the nominal or reference mean.
Candidates treat stability like a simple recovery calculation and forget that the reference is the time-zero or freshly prepared sample and that the ±15% criterion applies to the mean, with precision also required.
What to do instead
Set up the calculation explicitly: percent change = (mean stressed − mean reference)/mean reference × 100, evaluated against ±15%, using a freshly prepared curve for both sets, and report the precision alongside the bias.
The candidate answers a question about internal standard selection by choosing an isotopically labeled analog without addressing why a stable-isotope-labeled IS is preferred and what problem it solves, or conversely selects a structural analog and claims it corrects for ionization suppression equally well.
The rationale (co-elution and matched ionization behavior) is understood only superficially, so candidates cannot explain when a structural analog fails, for example when it elutes at a different retention time and experiences different suppression.
What to do instead
For every IS question, state the mechanism: the SIL-IS co-elutes and experiences the same matrix suppression, so the analyte/IS ratio cancels it; note that a structural analog only works if it co-elutes and is not itself subject to differential suppression.
The candidate computes a dilution integrity or dilution QC result by multiplying the measured concentration by the dilution factor but forgets to verify that the diluted sample's precision and accuracy are assessed against the undiluted nominal, and omits the requirement that the dilution not exceed the validated factor.
The arithmetic is done correctly but the acceptance framework is dropped; candidates do not recall that dilution QC must be prepared at a concentration above the ULOQ and diluted into the calibration range, with the same ±15% criteria.
What to do instead
Write the full workflow: prepare dilution QC above the ULOQ, dilute by the validated factor into range, back-calculate, and apply ±15% accuracy and ≤15% precision; explicitly flag that any factor beyond the validated one requires revalidation.
In a general analytical chemistry question on method validation, the candidate lists precision, accuracy, and linearity but omits specificity/selectivity, range, robustness, and detection/quantitation limits, or lists them without linking each to the experiment that demonstrates it.
Validation parameters are memorized as a list rather than as an experimental design, so under exam pressure only the three most familiar terms surface.
What to do instead
Anchor each parameter to its experiment: selectivity to blank and interference injections, range to the calibration span, robustness to deliberate small perturbations (pH, column temperature, flow), and so on; recite the parameter–experiment pair together.
When asked to interpret a chromatographic separation problem, the candidate adjusts only the mobile-phase organic percentage to improve resolution, without considering selectivity (α) versus efficiency (N) and the role of pH, column chemistry, or gradient slope.
The resolution equation is known as a formula but not as a diagnostic tool, so candidates default to the single knob they remember rather than identifying whether the problem is retention, selectivity, or band broadening.
What to do instead
Diagnose first: compute or estimate which term in R = (√N/4)(α − 1)(k/(1 + k)) is deficient; if α is near 1, change selectivity (pH, solvent type, column chemistry); if N is low, address kinetics (flow, particle size, temperature); if k is too small, adjust solvent strength.
The candidate reports a quantitative result with more significant figures than the calibration and instrument precision support, or propagates uncertainty by simple addition rather than in quadrature, and treats a result just inside a specification as unambiguously passing.
Significant-figure discipline and uncertainty propagation are typically learned early and then eroded by calculator output; candidates also confuse accuracy with precision when deciding whether a borderline result passes.
What to do instead
Round the reported value to the precision justified by the calibration and repeatability data, propagate independent uncertainties in quadrature, and when a result is within the uncertainty of a limit, state that it is inconclusive rather than passing.