Hi everyone,
I am trying to do a Differential gene expression analysis of a bulk RNA-seq. We have 4 different conditions : Control , G mutation, G and S mutation and G, S and A mutation (so it is kind of adding mutations to the already present). 2 replicates per 2 timepoints, making 4 samples per condition.
What we want is to extract the differential expressed genes of all conditions compared to control (not between them directly). So, for that reason, and also considering that the quality of the RNA-seq was not the best one, we analyzed included all the samples in the coundata matrix, using DESeq2 with Likelihood Ratio Test ('LRT), being the design=~batch+rep + condition and the reduced = batch+rep. When executing the results() function, we assigned there the different pairwise comparision with the contrast option (e.g. contrast = c("condition", "XX", "CTRL").
Up to this point, what is obtained after filtering for p-adjusted and log2foldchange it is obtained the differential expressed genes for each pairwise comparision. My questions are : With our objective, is that approach correct ? Would it be preferrable to use Wald test and pairwise comparisions? As the 'LRT' gives a p-value that is "independet" from the log2foldchange, up to which point we can also filter by log2foldchange and up to this point it makes sense to 'trust' this number?
It can be ensured that what we obtain are significantly differential expressed genes ? or it has other interpretation?
When we tried to plot a Volcano Plot, of course, we were not getting the V shape (the V gap in around FoldChange = 0) as of course the p-value is not as dependent to log2foldchange as it is in Walt Test. Would it be then this representation correct? is there any other way of representing?
As a final and general question, would you suggest any changes? do you see anything that does not make sense in our reasoning or anything that can be improved / better approached?
Thanks a lot!
