ELISA Data Analysis
ELISA is a plate-based assay used to detect the concentration of a target protein in biological samples such as peptides, proteins, antibodies and hormones. Three different types of data output can be acquired:
Qualitative: Simply obtain a yes or no answer to the presence of the target protein in the sample by comparing to the blank well or an unrelated control antigen without the target protein
Semi-Quantitative: Use the signal intensities to compare the relative levels of the target protein in the assay samples since signal intensity is directly related to antigen concentration
Quantitative: Precisely calculate the target protein concentration in assay samples by comparing to a standard curve with known target protein concentrations that have been serially diluted
Below, we discuss the different aspects to consider for more consistent and accurate ELISA data. Furthermore, we provide a step-by-step guide to create the standard curve for analysis.
Before running the ELISA
Before running an ELISA, we recommend performing some best practices:
- Read our blog on tips for preparing ELISA standards
- Make sure to run ELISA standards and samples in duplicate or triplicate so that the data can be statistically validated
- Calculate the average, standard deviation (SD), and coefficient of variation (CV) to determine pipetting precision. Although some labs suggest CVs of less than 20% for replicates, we recommend CVs for replicates to be less than 10%
- CV values greater than 20% could be caused by pipetting errors, contaminated plate or reagents, uneven temperature across the plate, or evaporation during incubation steps
- Generate a new standard curve for every plate to minimize the effect of operator, pipetting, incubations, and temperature differences
- Include a positive control sample with known concentration to help determine if the calculated concentrations of the samples are valid
- Run blank samples to compute the background absorbance
- Determine the appropriate dilution factor and dilute the samples so that the concentrations fall within the range of the standard curve
After running the ELISA
A standard curve (aka calibration curve) for the protein of interest is constructed by plotting the mean absorbance (y-axis) against the protein concentration (x-axis) and choosing the best fit curve for the data points. Based on the standard curve, we interpolate the sample absorbances to compute the sample concentrations. We will be discussing the steps in more detail below.
The Standard Curve
After obtaining raw data from the ELISA reader, the ELISA results are ready for statistical analysis. We suggest using an ELISA data analysis software for the analysis. Our lab works with CurveExpert 1.4, but many other curve fitting software and tools are available, such as GraphPad Prism. Microsoft Excel can also be used to analyze ELISA results, but it may not offer as many options or flexibility as other programs for scientists.
For this standard curve example, we will be using CurveExpert 1.4 to explain the process.
1. Enter ELISA data into software
Categorize the ELISA raw data into three sections:
- Absorbance of the blank well
- Absorbance of standards with known concentrations
- Absorbance of samples with unknown concentrations
It is important to run the blank well with sample diluent to determine the background absorbance. Even without the presence of the protein, the buffer will still have an OD value. The absorbance of the blank wells should be subtracted from all standard and sample absorbances for accurate OD readings.
Open “CurveExpert 1.4” to see the interface below:
Enter the standard concentration in the x-axis column and the corresponding OD values in the y-axis column. The data plot will be presented in the bottom right corner.
2. Select the best fitting curve
Click the [Run] button in the top menu bar to allow the software to examine the data and choose the best possible curve fit. The window below will show up:
Click [All On] to include all model families for calculation. However, if you prefer not to include all of them, specific model families can also be selected for calculation. If “Polynomials” is checked, you will be asked to input the polynomial constraint, which we recommend setting as “4”.
Press [OK] to run the calculation.
The resulting curve fits will be ranked based on the standard error and correlation coefficient. Double click on each model to see the corresponding curve.
Choose the curve that meets the following criteria:
- The equation with the higher R value
- The curve should rise smoothly and closely resemble a straight line
Right click and select “Copy” to paste the graph into an excel sheet or word document.
Our lab and most companies generally recommend using a 4-parameter algorithm for the best standard curve fit. [Why?]
In this example, we have chosen the quadratic fit curve. Apart from the polynomial fit, the quadratic curve has the highest R value and closely resembles a straight line that rises smoothly. [Why aren’t we using the polynomial fit curve?]
3. Calculate target protein concentration
The calculation can be performed in the software or with Excel. If the samples were diluted before the ELISA, make sure to multiply the computed sample concentrations by the sample dilution factor.
- Using software (CurveExpert 1.4) to find the sample concentrations
Using software will enable the user to easily find the x- and y-values, differentiate, and integrate the curve fit. Right click on the chosen curve fit graph for the graphing features menu and choose “Analyze”. For ELISA analysis, we would navigate to the “Find x=f(y)” tab and enter the sample OD value (y value) in the “At Y =” field. Click [Calculate] to obtain the x value (the target protein concentration) at the specified y value.
- Using MS Excel to find the sample concentrations
Click the [Info] button in the top left corner of the graph. This will provide the model information for the curve fit along with other statistical information for the model.
The “Coefficients” tab displays the model function and the values of the coefficients a, b, c, etc. Press [Copy] and paste the function and coefficient values in an Excel sheet.
In the Excel sheet, input the corresponding coefficient values and y values (OD value) into the formula to calculate the sample concentrations.
Why is the 4-parameter algorithm recommended as the best standard curve fit?
Curve fitting software will provide different model options for data plotting, including linear plots, semi-log plots, log/log plots, and 4- or 5-parameter logistic (4PL or 5PL) curves. Although linear plots with R2 values greater than 0.99 indicate good fitting, data points on the lower end of the range are compressed, which will reduce resolution. Semi-log and log/log plots resolve this issue. Data points are spread out more evenly with semi-log plots and log/log plots offer good linearity for the low to medium ranges of the curves. The 4- or 5-parameter logistic curves (4PL or 5PL) are more complex calculations that take into consideration additional parameters such as the maximum and minimum. The main difference between the 4PL and 5PL curves is that the 4PL curve is symmetrical around an inflection point, but the 5PL curve is asymmetrical. If the data points suggest asymmetry near the plateaus, the 5PL curve would be useful. However, more data points need to be collected to determine if asymmetry exists. As a result, it is generally recommended to use the simpler 4-PL for the best standard curve fit.
Why aren’t we using the polynomial fit curve?
Although the polynomial fit curve is ranked 1 in the list and the curve has the highest R value for the example above, we should avoid using the polynomial fit as the standard curve. One thing to keep in mind with polynomials is that data points may sometimes result in a fitted curve that reaches maximum OD and then goes down again. This will result in having two concentration values for the same OD value. For example, the 2 polynomial curves shown below are unsuitable to be used as the standard curves for your ELISAs.