ELISA analysis allows confirmation of the presence or concentration of an antigen in a sample.
Here is a guide to the different types of data available and best practices for using ELISA quantitatively via the calibration curve.
Drawing an ELISA Calibration Curve
The calibration curve is drawn by measuring known concentrations of a reference antigen and plotting them against the instrument readout using optical density. The majority of ELISA plate readers provide software for calibration curve drawing, fitting, and analysis.
The antigen concentration of a sample can then be determined by extrapolating the linear section of the calibration curve.
Figure 1: Example of a quantitative ELISA standard curve from Human ICAM1 SimpleStep ELISA® Kit (ab174445).
Data Fitting Models for Calibration Curve Software
data fitting gives the readout on one axis and the antigen concentration on the other. Statistical fitting is achieved using R2 values, with a value of 0.99 or higher showing a good fit. Linear plots, while useful, can result in decreased resolution as they can compress the data at the lower end of the curve.
data fitting is used to counteract the effect of the compression in linear plots. This is achieved by plotting the readout versus the log of the antigen concentration, which gives a more even data distribution in the form of a sigmoidal curve.
data fitting plots the log of the readout versus the log of the antigen concentration. This tends to give good linearity in the low to medium concentration ranges and a loss of linearity at higher concentrations.
4- or 5-parameter logistic (4PL or 5PL) curves
are a sophisticated data fitting method that includes parameters requiring further calculations, such as the maximum and minimum. The 4PL method assumes symmetry around the point of inflection whereas 5PL also accounts for asymmetry, which is more appropriate for this application.
If possible, use your analysis software to generate 4-PL or 5-PL curves, which will work with most ELISA results and provide the best data. If this is not possible, then a semi-log or log/log plot is recommended.