Hi

I want to perform DESEq2 vst and rlog variance stabilising transformations on gene counts from two different tissue types in two different mammalian species. I have two questions:

1. Is there a disadvantage to performing vst/rlog only on a subset of genes that are significantly different in my DESeq2 DE analysis? or should I be doing these transformations on the entire GeneCount matrices?

2. Does it matter if I perform these transformations using a matrix that has -

a. ALL genecounts from both tissue types in both species or

b. has counts for both tissue types in A species or

c. has genecounts from A tissue type in both species?

shall be grateful for any advice!

Thanks in advance, Adi

VST and rlog need to learn the global trend of dispersion over mean. I recommend to use vst() and to run on the entire dataset (all genes, all samples).