Has anyone ever worked with EBSeq-HMM?
As I know, EBSeq, uses "PP" and "FDR" to decide which genes are DE. For example under FDR 5%, a gene is said to be DE if its posterior probability (PP) of being EE (equally expressed) is <0.05. And if you want to choose most significant genes you can easily choose the ones with higher PP(DE) because we don't have a p-value or FC here to filter genes based on them.
EBSeq-HMM is developed for detection of DE genes in time course studies. So, for example, for a gene in 3 time points, the possible paths are as follows:
UP-UP (The expression path of the gene from Time 1 to Time 2 is "UP" and from Time 2 to Time 3 is "UP")
Using EBSeq-HMM, you are given a table of PPs for each possible path (remember that in EBSeq, you are only given PP(DE) and PP(EE)). Here, again, under FDR 5%, a gene is said to be DE if its PP(EE) is <0.05
Now the question is, when you use EBSeq-HMM, you don't have PP(DE). So how do you decide which genes are most significant? I know if I had p-values and FCs, I could use them, but here all we have is just PPs of all possible paths. And now that I'm using this package, I have a list of 16000 genes detected as DE which is way too many. How can I further filter them?
Please if my question is vague or you need more info, let me know.