Dear all,

we have been discussing internally but couldnt find an answer, maybe you could help us. The following happened: I run my DESEQ2 analysis with the full dataset and got genes with a certain log2fold change and pvalue via results(dds0). I then run the analysis again after removing some genes. We expected that the log2fold change and pvalue would not change and that only the padj value would change. But all values changed.. Can anyone explain why this is the case?

Thank you, Bine

Swbarnes is correct. The pvalue depends on the dispersion which borrows information across genes. Also the size factors could differ.