This is a test version of Biostars. For the public version, visit https://www.biostars.org.
RNA-seq with DESeq2 and different conditions

Hi everyone!

I am doing a deferentially expressed analysis with DESeq2, but now i have a doubt about using the contrast.

I have my info table of all samples that for example condition A, I have x, y,w,z different variables but I am only interested in analyse the deferentially expressed genes in y vs z, and what I do was using all samples and all condition, but I define the contrast=c("A","y",z") , but when i subset the info table and counts table for only samples and condition that i am interested and again i use the same contrast but this time the only variables is y and z, the results are different, that is in one analysis i have more genes than the other analysis. Someone can tell why this may happen and what I have to use, because I do not understand the differences.

Thanks for your help!

rna-seq r
• 1,824 views • link
written by martapatriciaribeiroferreira • martapatriciaribeiroferreira

1 answer

Why did you subset the counts table? Although the additional samples might not contribute to the contrast that you are interested in, they help estimate things like the average expression level and the dispersion-relationship. These things influence the model fitting and any downstream things like contrast calculation.

• 0 views • link
written by russhh russhh

Thanks for your reply! But I thought that aren't important use the samples that are not in my contrast and because I am doing a PCA and heatmap and I only can show in these graphs the samples of these two variables. So you thing that for a set of samples I should use always the complete matrix of counts but defining the contrast with only the variables that I want??

Thanks a lot!

• 0 views • link
written by martapatriciaribeiroferreira • martapatriciaribeiroferreira

Provided all the rest of your samples are of good quality.

• 0 views • link
written by russhh russhh

Log in to answer this question.

More posts like this