Hi! I have a question about limma! I have been using it to highlight interesting features in different MS datasets, I have been testing 2 groups and multiple groups and only considering at the end the features which result on having a small p-value (not gonna say significant!) after fdr correction, so far so good (I guess).
Let's come to my problem, I am currently analysing a dataset of controls and CKD patients and I have divided the data into four group: healthy (n=12), CKD stage 1-2 (n=35), CKD stage 3-5 (n=70), dialysis (n=29). After running the limma test and performing p-value adjustment, I get a list of 112 features, thing that I find quite curious because I have never got so many differently expressed features.
I also made a heatmap and from that a 4 cluster pattern is visible (and the clusters correspond mostly to the patients groups). But still I am worried that the test might have selected too many features which might also be redundant? I also add the fact that I have already previously filtered out the dataset and deleted features which had >80% missing values. Is the difference of the samples size of the groups influencing the result?
I hope my question was clear and that you can help me out with my doubts!