I did a cytokine quantification in human plasma with a custom procartaplex kit (thermofisher) on a Luminex xMAP (Bio-Rad). I have many samples that are below the detection limit (
instead of setting values to 0 (which will influence your statistics and likely is going to introduce an artefact, you may consider rerunning your samples with our 48-plex cytokine assay with absolute quantification. As it is a qPCR based readout, you have much better sensitivity.
You can check the independent comparison vs. Luminex here, in brief the PEA assay not only guarantees better sensitivity (and specificity due to dual antibody binding), but also a much broader dynamic range for different cytokine analytes, which should help especially when you have large variation within your experimental groups.
Thank you for this interesting proposal, however, we have no more funds for this project... I (finally) found an interesting article that proposes to use fluorescence instead of concentration. This way, I don't have any missing values and it even seems to be more powerful statistically.
Thank you for your answer. The main problem with this approach is that I have very little value for certain cytokines and I can't do good statistics ... And as I previously said, I have no more funds to reconduct the experiment and I have to do with this data...
I think that the use of fluorescence intensity is a good way in my situation, it's relative quantification but statistically powerful.
We recently ran into the same issue with ours as well.
We ended up using the standards to create a regression, then reverse predicting the actual values from the samples MFIs. This allowed us to report actual concentrations for the stats we needed.