Begin with plate and control behavior
Confirm blanks, positive and negative controls, standards, QCs, and unknowns are mapped correctly. Inspect raw absorbance, fluorescence, or luminescence before blank correction or normalization.
Review replicate variation and spatial patterns. Edge effects, dispensing order, incubation timing, washes, substrate development, and reader settings can create systematic behavior that a fitted standard curve will not reveal.
Fit the standard curve over a supported range
A 4PL is a common starting point for a symmetric sigmoid standard curve; a 5PL can represent asymmetry when repeated data support the extra parameter. Some methods use a weighted fit because variance changes across concentration.
Plot all standard observations and residuals. Review convergence, parameter plausibility, standard back-calculation, independent QC recovery, and behavior near the asymptotes. Define whether and how a standard may be excluded.
Back-calculate unknowns and apply dilution factors once
Invert the fitted curve to estimate unknown concentrations. Apply sample predilution and on-plate dilution factors with explicit units. Retain both the interpolated plate concentration and final adjusted result.
Values near flat asymptotes can be highly uncertain, and values beyond the validated range should be flagged according to the method. Reanalysis at another dilution is often more defensible than extrapolation.
Worked example: keep plate and final concentrations distinct
Suppose the fitted ELISA curve returns 14.2 ng/mL for an unknown well and the sample was prediluted 1:20. The preparation-adjusted concentration is 14.2 × 20 = 284 ng/mL.
Report the 14.2 ng/mL plate concentration, the factor of 20, and the 284 ng/mL final result. Apply any replicate summary and range rule at the stage defined by the method; do not hide an out-of-range well by averaging it with another dilution.
14.2 ng/mL × 20 = 284 ng/mLReport more than a standard-curve picture
Include plate and source identity, blank or normalization rules, standard and QC levels, model equation and weighting, parameter table, exclusions, back-calculated recovery, unknown replicates, dilution factors, range flags, final results, and settings and software versions.
A reviewer should be able to move from the final concentration back to the exact well response, curve version, inverse prediction, preparation factor, and range decision that produced it.
Frequently asked questions
Is 4PL always the right model for ELISA?
No. Select the model and range using representative assay data, residuals, recovery, precision, and the intended procedure. Some bounded ranges are linear; some curves show supported asymmetry.
Can I remove a failed standard point?
Only under a predefined, justified rule. Preserve the value and reason, assess the effect on curve and range, and report the exclusion.
Should unknowns be averaged before applying dilution factors?
Follow the defined method and calculate from unrounded values. Preserve each replicate, plate-level interpolation, factor, and final aggregation rule.