What are your assay conditions? I have had good success determining the quantity of murine mAbs in "conditioned" supernatant samples using ELISA and a standard curve generated using the appropriate IgG subclass. Is the antibody purified, or is it in serum or culture supernatant? Are you trying to quantify the concentration of a mAb or pAb?

Hey Hanif, is this in serum or is it a monoclonal? I have lots of experience with serum antibody titres, and we're at the same university! Let me know if you think I can help

Hi Greg. thanks for your answer, it is purified monoclonal antibody, my antibody is called g-12 Flightless antibody. i am trying to quantify mab concentration. I have determined that ELISA only picks the concentration difference from 10 ug/ml. but i still cant get a linear standard curve.

Hi Natalie

my antibody is monoclonal, i have tried few optimisation but still cant get a linear standard. Any suggestions would be helpful. 

What are you using to quantitate? Anti-mouse antibodies for coating and detection? Or antigen for coating? In principle these assays should be possible.

Hanif competition ELISA normally give a good inverse relationship between OD and mAb concentration. Put your data in Excell and then use the Excell analysis to draw a trendline.  Check R (given by the program). I have found that most of my competition ELISAs gave me a inversely proportional logarithmic curve.  For DAS I get an area with linearity. However at very high concentrations reaches a plateau and sometimes even get dip/prozone.  Results obtained depends on how indirect ELISA is done. 

You might consider sharing your data so that we can better understand the issue you are experiencing

Agree with Greg.  Also the exact method used ie. coating, detection, conjugate.

Maybe your antibody is not one clone.  I would try a competitive assay.

i have attached the protocol and some results. is there any optimisation that can be done to overcome the saturation of the antibody at higher concentration. 

Hi Hanif, again I mostly perform polyclonal assessments, but I do use HRP and OPD to develop. Here are some pointers to try to see if you get improvement:

It looks as though your peptide coating concentration is 40ug/mL -  I have worked with full proteins, not peptides, but to me this seems like heaps! I would routinely use 2-5ug/mL of purified protein for coating. 10ug/mL would be a good starting point.

I also generally block first without tween, I think it enables more binding to the plate and potentially reduces background. Though as your BSA is so higly concentrated you should still get high binding.

I think your antibody concentration is too high. I usually start at around 100ng/mL for the top of my standard curve, and do 1 in 2 or 1 in 4 dilutions, and this can get maxed out even for the top 2-3 wells. If you start from a much lower concentration you can develop the plate for a little longer (I dont have a set time, I wait for the start of colour to develop in the blank wells, or until all my standard wells show something, and then I stop). If you do closer dilutions your quantitation will be much better. For serum I was routinely diluting the samples 1:50,000 to get a good concenration, using 5ug/mL of coating protein.

I analysed your data and displayed it with a log(x) axis. It looks like you have acheived a linear curve between the concentrations of 10-500 ng/mL (0.01-.5 ug/m). I guess the question is do you need an ELISA that can detect several orders of magnitude difference in concentration? If your samples really could be that different you could always run the ELISA with a standard curve from 1-500 ng/mL and just run a couple dilutions of your samples (say a 1 in 10 and a 1 in 100) as one is likely bound to be in range.

Also I noticed you are reading at 450. With OPD I would always 'stop' the colour change reaction with the addition of acid (50uL 3M HCl per well) and the colour goes orange. Then I would read at 495. I found I got more sensitive detection this way.

I made a graph! Note the log axis on the bottom. You can see the linear relationship here, and you can use nonlinear regression to fit a cuve to the transformed data

Thank you Natalie for taking your time to help me out. I have been running few other assays but still struggling to get linear standard curve. I have tried narrower range of concentrations but couldnt achieve what i was demanding. Here is the ELISA standard curve. 

Most Elisa's curves can only accurately cover a log.  Also I have never seen a linear curve, I always use best fit from excel..  

thank you judith, are you saying that using logarithmic curve is fine ? 

Hi Hanif, I think using a logarithmic curve is standard practice. This will be fine for your methods (estimating antibody concentration) but I wouldn't go over a log or two in concentration for the curve.

So it means your ELISA worked!

Once again thank you Natalie. unfortunately, for my anlytical method i need to generate a reasonable standard curve with high R value. I can use the logarthmic curve to quantify my samples but it would cause problems down the track and would certainly raise questions in my thesis. 

Hi Hanif,

there are several methods to get an idoea about the quantitiy of the antibodies in your sample. You have to take into account, that two parameters influencing the signal in an ELISA: 1. the affinity (or in a polyclonal system, the respective mixture) and 2. the concentration.  

In General, you need to have (define) a limit from where on a test result should be interpreted as “positive”. There are several measures:

Titer, concentration, affinity, …

S/C (sample/cut-off-Ratio)

For the detection of a titer you need to have two things:

·         A serial dilution of your sample and

·         A point / measurement value from where you define a reaction as positive.

This could be the blank value / negative control + 2.96 fold standard deviation or anything else

If you would like to work without serial dilutions (serial dilution requires much more material and costs to measure a single sample) only with a single sample dilution, you need to have also a point from where on you define a sample as positive (or if you using calibration curves to determine a concentration, this point is a concentration limit). It’s always the way from your measurement values (OD, in RIA cpm, in CLIA RLU, …) using some test standardization procedure (negative control, calibration curve, titration, …) into a much more usable form.

Using these limits is quite usual and helpful for the interpretation of the results, i.e. to dichotomize your results in “positive” or “negative” (forget all the discussion about a grey zone, even here you have to decide what to do). In general, this dichotomization is possible for OD’s, RLU, CPM, titers, etc. ... even for S/CO or P/N-ratios. It’s always nothing else than a way to get the system standardized by introducing controls or standard (calibrators). See attached figure “EIA_Calibration_curve” and Folie1”. No 1 – 2d are dealing with the calculation of the results, only 3 with the cut off.

The determination of the cut off (for OD’s, RLU, CPM, titers, etc. ... even for S/CO or P/N-ratios) depends on your question you try to answer with your assay. Let’s do two examples:

Using an assay for the detection of HIV in a blood service environment you need to detect as much as possible infected persons. In statistical terms, you need a high sensitivity. Just a short stop: How to get this: Define simply everyone as positive so will would never overlook someone. The threat of this, no one is negative and no one could donate any blood. In statistical terms: it’s a very low specificity. Here you can see already, that both Sensitivity and specificity have to be always taken as a pair. Never alone.

The other example: Cancer screening using PSA (prostate specific antigen). There are many reasons to get the “tumor marker” increased. Bicycling is already sufficient for this. To avoid frightening to many men, you should use a high specificity and you can live with a low sensitivity.

Concluding this, you realized already that you need to have two populations, which are defined by other methods than your ELISA as “positive” or “negative”. This other method could a clinical diagnosis, a radiological method, an alternative assay or test. Now you are measuring sample from both populations and compare the results, either as OD’s, CPN, RLU, titers,… and you will get a distribution of you measured/calculated values for the “positive” and the “negative” population. Now you have to search for the measurement value (OD’s, RLU, CPM, titers, S/CO or P/N-ratios, …) which allows you the best discrimination between this two populations. So, you should summarize your results in a 4 filed table.

                     Diagnosis positive    Diagnosis negative   Sum

Test positive   right positive            false positive            all test positives

Test negative  false negative           right negative           all test negatives

Sum               all positive               all negative

[SE] Sensitivity = all positive / right positive

[SP] Specificity = all negative / right negative

And now, the index Youden introduced (http://www.ncbi.nlm.nih.gov/pubmed/15405679, see also the attached paper). In figure 1 we showed the distribution of the measurement values of the two populations. Youden simple added Y=(SP + SE)-1. So you get only one curve above the value (OD, CPM, RLU, titer, …) showing an optimum. At this point is the optimal cut off if SP and SE are not weighted. If you have some prerequisites for a higher SP you have to introduce in the formula only a weighting factor. Y=(a*SP + b*SE)-1, whereby a + b = 1.

Summarizing all this: OD, Cut of, S/CO, P/N, Titer … this are ways to standardize your assay. The cut off is for the evaluation; interpretation of the test results.

The only problem is, that you are using your positive and negative test population for the optimization of your assay. It’s a “training population”. To make sure, that your test is valid in reality as well, you need to have a “validation population” as well. And that’s a huge amount of work.

At the end, you will see, that the value, even if it’s below the detection limit, isn’t important anymore. By using cut off’s you are switching from an analogous scale to a dichotomic test results (there is only “positive” or “negative” left). Again, there are two issues to solve:

1.    The assay standardization to make sure that you are measuring at each assay, at each test day, at each test lot … the  same

2.    The test interpretation using the cut off.

I hope, that I explained it well. If you have any question, just ask me.