Entering edit mode
5.8 years ago
roy.granit
▴
890
I ran an mRNA seq analysis with three groups: control, treated1, treated2 - each with two replicates.
When I look at the comparison of 'treated2 vs control' I notice that most of the genes that hit statistical adjusted p-value are 'log2FoldChange' positive - meaning they are up-regulated in treated2.
Does this make sense? I guess I would expect to see relative equal numbers of up/down regulated genes.. what can you make of this?
Thanks
That sounds like more of a biology question than a bioinformatics question. Based on the treatment would you expect global/general upregulation? Some treatments cause global gene silencing/upregulation so those results would be expected.
Are you also applying a log2FoldChange cutoff? If you apply that cutoff (look for genes with |log2FC| > 1 and FDR < 0.05) do you still see this pattern?
To add to the above, it would also be useful to know if you are applying a minimum count cutoff per gene? Perhaps if you are applying a cutoff well above the default, it could be removing a certain number of genes (up and down-regulated), although of course removing low-expressing genes also has an effect on FDR adjustment. Just a few more details would be helpful.
roy.granit, the point by lshepard is quite valid. Indeed, nobody can really answer your question because you have not given a great amount of detail about your entire processing pipeline. It is also important to know how you derived the raw counts that you input to DESeq2.
Thanks everyone. I'm not applying any cutoffs other than FDR < 0.05.
As for the details, this is 3' RNAseq lib which I have aligned to the genome/transcriptome using STAR and counted the reads with Salmon.
My main concern was that this is caused by some mistake in the analysis pram, but since in this case I believe that can provide a reasonable biological explanation.
Okay, cool.
How does the MA plot look like?
Looks pretty equal..
Then you should be fine. Many times you only get genes up or down regulated