I have a different question but to the same topic, so I hope i don't need to start a new question.
I have analysed the same data now both with the two-step normalization using DESeq2 but also with the erccdashboard package.
In the DESeq analysis i have got almost 400 genes with an adjusted
p-value<=0.05. Not one of these genes has a
This is because most of the genes with a positive log2FC has 'NA' in the padj column (~90 from ~6000 genes with log2FC>0 are !='NA' in the results list. All the other genes are NA).
If I am doing the DESeq2 analysis without the independent filtering (by setting it in the
results function to
FALSE), I am getting only 12 genes with an adjusted
As I was wondering why that is, I have also tested the erccdashboard with its QuasiSeq normalisation and differential expression analysis procedure. Here I am getting quite a different picture.
I have "only" 1720 genes with significant adjusted
p-value <=0.05, but around 50% are positive (
log2FC>0). Most of the statistically significant down-regulated genes are common to DESeq2 and erccdashboard, but none of the up-regulated genes in QuasiSeq.
Is there a reasonable explanation for such behaviour?
How come QuasiSeq calculate a significant p-values for different genes? Are the p-values in the erccdashbaord result files being adjusted for multiple testing hypothesis?