I am using maSigPro to analyse a time serie data. I am very new to this kind of analysis and I would have a few (begginer's?) questions about the software and the result's interpretation.
First, I tried to find an indication on what would be the most appropriate software to normalize the data if we want to use it in maSigPro? (many scientists use EdgeR before running maSigPro but I am currently using DESeq2 as I already performed the pairwise comparison with it and as I would like to see the interesct between the DEG list over time and at a specific time point).
Then, when I use the "p.vector" function for RNAseq data (NB model) and then I fit the model with t.fit, I get :
1 754 influential genes with paramater "step.method = forward" 1 146 influential genes with paramater "step.method = backward" 1 755 influential genes with paramater "step.method = two.ways.forward" 1 126 influential genes with paramater "step.method = two.ways.backward"
My second question is : After checking these genes some have a lot of zero still (even if I already applied a pre-filter) and some others have, in two samples only (different condition but same repetition), a high expression. Is it necessary to remove all these influential genes from the data before re-running maSigPro ? How does it* influence the rest of analysis/DEG list if I don't*?
My other question is : why was there such a big difference between all these step methods ? And according to you which one is the most reliable (if it is possible to say of course)?
Thank you in advance for your help.
PS: I hope someone can I have a beggining of answer.