Hi guys,
I am trying to find the best approach to assign a value to samples that do not reach the baseline value in ELISA.
In my ELISA assay I perform a 2-fold serial dilution curve starting at a 1:200 dilution. Next to a pos, neg and blank control I run a titer control sample in every plate that results in an final O.D. of about 0.1. I compare all the serial dilutions of every sample and look for the highest dilution that still has an O.D. above the O.D. of the titer control sample. This allows me to get a "titer" for every sample, and these results are comparable with results obtained from different ELISA plates because they were compared to the same control sample.
Now sometimes the O.D. in samples is already so low at 1:200 that it doesn't reach the O.D. of the titer control sample. A little example:
Sample A) 1:200 0.05, 1:400 0.05, 1:800 0.05
Sample B) 1:200 0.14, 1:400 0.07, 1:800 0.05
Sample C) 1:200 0.18, 1:400 0.09, 1:800 0.05
Sample D) 1:200 0.3, 1:400 0.15, 1:800 0.075
Titer control sample: 0.15
So it is clear that sample A and B do not reach the titer cutoff, and that sample C and D do (and will have a titer of 200 and 400 respectively). I am wondering whether I should assign sample A and B a baseline titer of 200, or whether I should assign something else. When reading about Limits Of Detection (LOD), people talk about assigning values like LOD/2 or LOD/Rootsquare(2), but I dont think I am dealing with a LOD in this case.
Please let me know if you have suggestions :)
I dont understand your data. Sometimes the OD increases at higher dilutions, is that right? Sample B has a titer in my opinion. Why do you say it doesnt?
Anyway, if no OD above the limit can be detected, I would say the titer is lower than 1/200. This is a fact. You would thus avoid to assign a titer that you cannot measure.
Sorry Gertrudis, I made a mistake with the decimals in my example. I corrected it now.
Regarding your answer, if you just say that your titer is lower than 1/200, how do you perform statistical analysis then? In my case I would like to include all the samples in my ANOVA-analysis, but leaving out data results will affect my statistics. How would you deal with this?
Avoid making serial dilutions out of the range of detection of your instrument, otherwise your data will be meaningless.
Actually I havent done this kind of statistics with titers. Assuming that non-detectable responses correspond to 1/200 is a posible trick, of course, but if you do that, you are somehow modifying the reality because you found no titer... Could you test lower dilutions to try to find the real limit?
Another possibility, although technically difficult, is to increase your assay sensitivity in a way that the number of non-detectable responses is lower. If all the titers increase with a more sensitive assay, the ratio between them should be conserved.
I can't test lower dilutions because I don't have enough serum to do that. Usually the samples that do not reach the 1:200 dilution titer cutoff are negative control samples, but I still want to assign a value to them, for statistical purposes.
For negative controls probably you wont be able to mesure any response. Perhaps its better not to use statistics. Assuming a titer for negative controls (which you ser not able to measure) is not very realistic You could find another way to present your data.
If half of your samples are below the limit of detection, you need a better assay. This one is obviously not working for you. What is your "titer control"? Is that your background?
Thanks Gertrudis, I will look into other ways to present/analyse my data.
Bernard, my titer control is a sample that was chosen to compare the unknown samples to. This titer control will always be the same between different ELISA plates and allow me to compare unknown samples from plate A to unknown samples from plate B. The background is already substracted in the values I show.
We always compare low samples to the background, if their values are not 3 times the background they are negative, but we repeat them with a more concentrated sample.
- Badges
- Science method
- Similar topics
- Medicine
- Immunology
- Immunology Techniques
- Immunoassay
- ELISA