Hi,

I noticed in the DESEQ2 tutorial (https://bioconductor.org/packages/release/bioc/vignettes/DESeq2/inst/doc/DESeq2.html#data-quality-assessment-by-sample-clustering-and-visualization) one doesn't standard normalize VST transformed data prior to PCA and/or clustering. Is it not advisable to do so? The VST eliminates the mean/variance relationship but still results in certain genes having a lot more variance (in transformed values) than most of the other genes.

Thanks,

Owen

The genes with high variance are typically those in which the samples are different for biological reasons, eg differential expression. If you squash all the genes to have equal variance you demote the biological signal and promote the noise. I’ve never understood why people suggest to do this.