Sandwich Enzyme-Linked Immunosorbent Assay (Sandwich ELISA) is a common method for detecting the concentration of
specific antigens in samples. Its data analysis focuses on establishing a "concentration-signal value" correlation model
using standards, then back-calculating the antigen concentration of unknown samples. The process involves four key steps:
data preprocessing, standard curve construction, concentration calculation of unknown samples, and result validation, as detailed below:
Blank control correction: The OD value of each test well should be subtracted by the OD value of the "blank well" (only filled with buffer and enzyme-labeled reagents, without antigen and antibody) to eliminate non-specific signal interferences such as reagent background and instrument errors.
Example: If the OD value of a sample well is 1.2 and the OD value of the blank well is 0.1, the corrected OD value = 1.2 - 0.1 = 1.1Duplicate well data processing: 2-3 duplicate wells should be set for each standard concentration or unknown sample. Calculate the mean (Mean) and standard deviation (SD) of their corrected OD values. If the OD value of a duplicate well deviates from the mean by more than ±10%-15% (adjustable according to experimental requirements), it is identified as an "outlier" and excluded to avoid the impact of random errors on results.
Negative/positive control validation: "Negative controls" (samples known to be free of the target antigen) and "positive controls" (samples known to contain the target antigen) must be included. If the corrected OD value of the negative control is too high (close to the OD value of the low-concentration standard well) or the OD value of the positive control is too low (failing to reach the expected signal), it indicates potential issues such as contamination or reagent inactivation in the experiment, and the experiment should be repeated.
Standard concentration gradient setting: Dilute the target antigen standard into 5-8 gradient concentrations (e.g., 0, 1, 5, 10, 50, 100, 200 ng/mL), which should cover the expected concentration range of unknown samples and include a "0-concentration" standard as a blank calibration point. Detect the corrected OD value of each
concentration.
Fitting model selection: Based on the binding characteristics of antigen and antibody, two common models are used:
Linear Regression: Applied when the corrected OD value shows a significant linear relationship with the concentration (correlation coefficient R² ≥ 0.98) within a certain range of standard concentrations (usually the middle concentration range). The fitting formula is y = ax + b (y = corrected OD value, x = antigen concentration, a = slope, b = intercept).
4-Parameter Logistic Regression (4PL): When the standard concentration range is wide and the OD value rises rapidly
first and then levels off (showing an "S"-haped curve) as the concentration increases, linear regression will have large errors.
The 4PL model should be adopted, with the formula y = (A - D) / [1 + (x/C)^B] + D (A = upper limit of OD value,
D = lower limit of OD value, B = slope factor of the curve, C = midpoint concentration (EC50), x = antigen concentration, y = corrected OD value). The 4PL model can more accurately fit non-linear relationships and is the preferred model for
sandwich ELISA (especially for high-sensitivity detection).
Standard curve validation: Evaluate the curve quality using the correlation coefficient (R² ≥ 0.98 for linear regression, R² ≥ 0.99 for 4PL) and residual analysis (deviation between the measured OD value of each standard concentration and the predicted value of the fitting curve). If R² is too low or residuals are too large, re-optimize the standard dilution gradient or experimental
conditions (e.g., antibody concentration, incubation time).
Concentration back-calculation: Substitute the corrected OD value of the unknown sample into the fitting formula of the standard curve to calculate the "detected concentration" (i.e., the concentration of the sample in the test well).
Example: If the corrected OD value of an unknown sample is 0.8, and the 4PL formula of the standard curve is y = (2.0 - 0.1)/[1 + (x/20)^1.5] + 0.1, substituting y = 0.8 gives a detected concentration x ≈ 12 ng/mLDilution factor recovery: If the unknown sample (e.g., serum, cell lysate) is diluted before the experiment (e.g., 1:10 dilution) because its concentration exceeds the standard curve range, the final concentration = detected concentration × dilution factor (in the above example, the final concentration = 12 × 10 = 120 ng/mL).
Concentration range judgment: If the corrected OD value of the unknown sample is lower than the OD value of the lowest concentration point of the standard curve, the result is labeled as "< lower limit of quantitation (LLOQ, e.g., < 1 ng/mL)". If it is higher than the OD value of the highest concentration point of the standard curve, label it as "> upper limit of quantitation (ULOQ, e.g., > 200 ng/mL)". Extrapolation is not allowed, and the sample should be re-diluted for detection.
Precision validation: Including "intra-assay precision" and "inter-assay precision":
Intra-assay precision: In the same experiment, detect a quality control sample (standard or reference sample with known
concentration) 6-8 times repeatedly, and calculate the coefficient of variation (CV% = (SD/Mean) × 100%). Generally,
CV% < 10% is required.
Inter-assay precision: Repeat the experiment for 3-5 consecutive days, detect the same quality control sample,
and calculate CV%. CV% < 15% is required.
Accuracy validation: Evaluate using "spike recovery rate": Add a certain amount of standard (low, medium, and high spiking levels) to a sample with known concentration. Calculate the recovery rate = [(measured concentration after spiking - background concentration of the sample) / spiked concentration of the standard] × 100%. The ideal recovery rate ranges from 80% to 120%, indicating that the detection result is accurate and there is no significant matrix interference (e.g., proteins or ions in the sample affecting antigen-antibody binding).