[BioC] voom

[Ina Hoeschele]

Hi,
  I'm using limma's voom function on a (human) RNAseq dataset (regular transcript based count matrix). When I do an MDS plot on the input count matrix, all looks "normal". When I do the same plot on the output log CPM matrix (E), the samples (N=329) cluster strongly into two groups of about equal size. This happens no matter how I run voom (with or without design matrix, passing count matrix and adjusted lib sizes, or DGE obj).
Thanks, Ina