Always run ELISA samples in duplicate or triplicate. This will provide enough data for statistical validation of the results. There are several pieces of software available to help process ELISA results in this manner. Calculate the average absorbance values for each set of duplicate standards and duplicate samples. Duplicates should be within 20% of the mean.

ELISA analysis

ELISA assays can be classified as follows according to the type of data obtained:

• Qualitative ELISA only determines whether the antigen is present or not in the sample. It requires a blank well containing no antigen or an unrelated control antigen.
• Semi-quantitative ELISA allows the relative comparison of the antigen levels between the samples.
• Quantitative ELISA allows calculation of the amount of antigen present in the sample. It requires comparison of the values measured for the samples with a standard curve prepared from a serial dilution of a purified antigen in a known concentration. This is the most commonly reported ELISA data.

Standard curve

The standard or calibration curve is the element of the quantitative ELISA that will allow calculating the concentration of antigen in the sample. The standard curve is derived from plotting known concentrations of a reference antigen against the readout obtained for each concentration (usually optical density at 450 nm). Most ELISA plate readers will incorporate a software for curve fitting and data analysis. The concentration of the antigen in the sample is calculated by extrapolation of the linear portion of the standard curve.​

Create a standard curve for the target protein by plotting the mean absorbance (y-axis) against the protein concentration (x-axis). Draw a best-fit curve through the points in the graph (we suggest that a suitable computer program be used for this). We recommend including a standard on each ELISA plate to provide a standard curve for each plate used. A representative standard curve is shown in the figure below from the human HIF1 alpha SimpleStep ELISA kit® (ab171577). Each point on the graph represents the mean of the three parallel titrations. We recommend using a sample of known concentration as a positive control. The concentration of the positive control sample should be within the linear section of the standard curve in order to obtain valid and accurate results.

Figure 7. ELISA standard curve example.

Curve fitting software allow using different models to plot your data.

• Linear plot presents the concentration of the antigen in one axis and the readout in the other. R2 values are normally used here to determine fitting, with values higher than 0.99 representing a very good fit. However, linear plots tend to compress data points on the lower end of the curve resulting in decreased resolution.
• Semi-log plot helps counteracting the compression at the lower end caused by linear plots. Semi-log plots use the log of the concentration against the readout. This method commonly results in a sigmoidal curve that distributes more evenly the data points.
• Log/log plot provides good linearity for the low to medium range of the concentrations. The higher end of the range tends to lose linearity.
• 4- or 5-parameter logistic (4PL or 5PL) curves are more sophisticated methods that take into account other parameters such as maximum and minimum and therefore require more complex calculations. 4PL assumes symmetry around the inflection point while 5PL takes asymmetry into account, which normally is a better fit for immunoassays.

If your software allows it, 4-PL and 5-PL will fit most ELISA calibration standard curves. If not, the best option is to use a semi-log or a log/log plot.

Competitive ELISA standard curve

For competitive ELISA, the antigen concentration is determined from the standard curve in the same manner as a conventional ELISA. The only difference is that the standard curve is inverted, with the highest concentration corresponding to the lowest OD value and vice versa, as seen below.

Figure 8. Competitive ELISA standard curve.

For competitive ELISA, the binding ratio, B/Bo, is a useful metric for optimizing the range of the standard curve. To He the binding ratio, simply divide B, the OD at a given concentration of analyte, by Bo, the OD when no sample analyte is present (i.e., the maximum OD for the assay.) If the binding ratio for a point on the standard curve is lower than 5% or higher than 95%, consider shifting the standard range as this may improve the interpolated values at the high or low ends of the curve.

Concentration of the target protein in the sample

To determine the concentration of target protein concentration in each sample, first find the mean absorbance value of the sample. From the y-axis of the standard curve graph, extend a horizontal line from this absorbance value to the standard curve. For example, if the absorbance reading is 1, extend the line from this absorbance point on the y-axis (a):

At the point of intersection, extend a vertical line to the x-axis and read the corresponding concentration (b).

Samples that have an absorbance value falling out of the range of the standard curve

For samples that have an absorbance value falling outside of the range of the standard curve, to obtain an accurate result, these samples should be diluted before proceeding with the ELISA staining. The concentration obtained from the standard curve when analyzing the results should be multiplied by the dilution factor.

Calculating the coefficient of variation

The coefficient variation (CV) is the ratio of the standard deviation σ to the mean µ:

C_v= σ/µ

This is expressed as a percentage of variance to the mean and indicates any inconsistencies and inaccuracies in the results. Larger variance indicates greater inconsistency and error. Some computer programs can calculate the CV values from ELISA results.

High CV can be caused by:

• Inaccurate pipetting; ensure pipette tips are sealed to the pipette before use so they draw up to correct volume of liquid
• Splashing of reagents between wells
• Bacterial of fungal contamination of either screen samples or reagents
• Cross contamination between reagents
• Temperature variations across the plate; ensure the plates are incubated in a stable temperature environment away from drafts
• Some of the wells drying out; ensure the plates are always covered at incubation steps

Spike recovery

Spike recovery determines the effect sample constituents have on detection of the antigen by the antibody. For example, the many proteins contained in tissue culture supernatant may hinder antibody binding and increase the signal to noise ratio, resulting in underestimation of the target concentration.

Known concentrations of protein are spiked into both the sample matrix and a standard diluent. The spiked protein is quantified using the assay and results from the sample matrix and the standard diluent are compared.

If the results are identical, then the sample matrix is considered to be valid for the assay procedure. If the recovery is different, then components in the sample matrix are interfering with the analyte detection.

What if a spike recovery experiment indicated that the sample matrix is affecting the results?

We recommend producing the standard curve using standard diluted in the sample matrix. Any effects on the results from the sample matrix will also be present in the standard, and therefore comparison between the standard curve and the samples is more accurate. Many of our ELISA kits contain a standard serum diluent for this purpose.

Another solution is to alter the sample matrix. For example, if neat biological sample is used, try diluting this in standard diluent. However, with this option, you will need to ensure that the dilution factor is taken into account when analyzing the results and that the concentration stays within the linear section of the standard curve.

View more of our ELISA kits and protocols or review our membrane antibody arrays, such as cytokine array ab133997, which can be used to measure many proteins simultaneously.