I am working with a competitive ELISA to detect and quantify a pesticide dissolved in water from one of two treatment types. If I were not particularly interested in the exact quantities in each treatment, rather, whether there was just a difference or not between the treatments, would it be acceptable to do a t-test on the raw OD's and not convert them using a standard curve?
Hi there,
There is no point comparing directly raw data from different experiments. From the same sample used in different experiments variation in raw data may arise just for technical reasons (level of saturation, adsorption efficiency of the antibody, antibody titer aso...). In summary, raw data have to be processed using controls and calibration curve from the same experimen prior to any comparison with data from other experiments.
If all your samples are in the same eLISA plate and tehre is no saturation (your samples are clearly below the values of other controls or samples), you can compare them.
If you are talking about different ELISA experiments there is no way to compare without a standard curve.
I am talking about all samples being run on the same ELISA plate. I have 20 true biological replicates of each treatment. Almost all values of one treatment fall just below the lowest value of the standard curve, while almost all values of the other treatment fall just above the lowest value of the standard curve (but inside the standard curve, nonetheless). It made sense to me to just analyze the raw OD's because all I care about is whether the treatments are different or not.
I should add that we also conducted replication on a positive control and a negative control, so we know the ELISA kit worked.
Yes, I think you can compare them. But conclusions should be carefully anlyzed, as your values seem to be around the detection limit of your assay. A higher sensitivity assay would be better.
I would apply a Mann-Whitney test on such data (over a normal t-test) as ODs are not really quantitative without your standard curve + plus ODs below LOD are not very reliable.
If all needed controls are included and samples are on same plate, then I see no problem using the raw ODs.
The biologic relevance of your data will be a bit more difficult to discuss - there might be a significance difference, but is this difference of biologic importance (may not easy to tell with ODs)?
- Badges
- Science method
- Similar topics
- Medicine
- Immunology
- Immunology Techniques
- Immunoassay
- ELISA