Possible to calculate avidity/affinity from an ELISA? - (Jan/06/2012 )
I'm comparing 2 different sets of antibodies to see which is more effective at binding a particular protein.
I've done an ELISA and got 2 curves showing one IS more effective than the other at both high and low concentrations of antibody by measuring optical density, but I'm having difficulty quantifying the difference. Is it possible to calculate avidity/affinity/some proper measure directly from optical density results? Neither curve has reached a plateau so I have no max value.
If not, is the best I can calculate simply a relative % difference between optical densities of the 2 antibodies? I would really like to quantify them in a way not necessarily relative to each other if possible.
Also, if one antibody fails to bind at low concentrations, but the other antibody DOES bind at low concentrations - is it possible to distinguish between avidity and affinity and say that this reflects AFFINITY because concentrations are so low that that is the sort of interaction which is most likely? Or would it still be impossible to distinguish avidity/affinity?
I've never had to analyse this sort of data in any depth before except for the obvious "one is better than the other", so I'm a bit shakey on how the data correlates with concepts I know already (affinity, avidity etc.).
Any advice would be appreciated!
In the ELISA plate we are looking at binding in that setting...with an immobilized antigen that is distorted by being bound to plastic, and the readout obtained using a secondary detection reagent some time after the antibody has been removed from the antigen so there is opportunity for dissociation. This may be quite different to the binding kinetics in solution. If you have a competitive format, where the antibody is anchored via the Fc and you are working with labelled and unlabelled antigen, your readout may be more real time if your final wash is done cold.
If you make fabs from your antibodies (which I presume are both IgGs) and your most effective antibody is still the most effective by about the same amount as for the intact assay, then the relative affinity is driving the OD difference you see. If you get entirely different ratios with the two fab preparations than you see with the intact moleclues, then avidity may be a factor in the OD differences.
I don't know if this is of any help
Hey, thanks a bunch for your reply!
Unfortunately, though, I can't do any further experiments. The thing I'm struggling with is writing up and analysing the data I already have. We've got a list of questions to go through which hint very heavily that there is some way of quantifying the differences between the two curves I've got, so that's where I'm stuck.
I've been trying to google it but only found information similar to what you've provided me with - further experiments.
OK, so I'm presuming that you have data from indirect ELISAs where the antigen is on the plate and you are generating a curves by varying the concentration of the test antibodies which are then detected using labelled anti antibodies.
We know that the experiments were designed not with the goal of quantfying and differentiating differences between affinity and avidity and we know that the data will not yeild accurate assessments of either.
I suspect what the reviewer is suggesting is that you do some non linear regression approach to derive affinity parameters from the data you have. This is fine to do, but if you do that, you will have to present the analysis highlighting the limitations of the approach which is the possibility that one antibody binds with one arm, the other both, or both antibodies with both, and that the antigen may not be presented as it would be in solution, and that the binding interaction was not stopped and the results not assessed with the minimum of delay because no cold wash was performed and a subsequent incubation which allowed an undefined level of dissociation to occur. Additionally, if one of your antibodies is IgG1 and the other IgG2 or some other variant for example, you have the considerations that the binding of the detection antibody to your test antibodies may also vary.
That was extremely, extremely helpful, thanks so much!
Now, off to Google exactly what non-linear regression is and how to use it to derive affinity parameters...
Daily I wonder at what missed stage of my education somebody was meant to have taught me all this stuff. I'm starting to wonder if I spent half a year in a coma at some critical point of my life and whilst everybody else was learning that a log is not a piece of wood (but instead some arbitrary and random thing you to do numbers on to get some other numbers for no apparent reason) I was in a dream-like sleep. Not knowing or having heard of these apparently 'basic' things has left me so incompetent it's like a disability. Just had to have a mini-rant here because it really frustrates me. Absolutely no reflection on your advice, just on my inability to act on it...
Ligand binding kinetic predictions by experimentation is a science in itself.
From what I can tell, Graphpad is a software that is commonly used in non-specialist labs so you might be able to get some numbers out of your data using it. You might even be able to get a free demo version.