I have gene expression (raw count and FPKM) from healthy and disease states. I'm going to do supervised and unsupervised classification on gene expression. As dimension reduction I'm going to get differential expression genes from DESeq2 and use these genes as features in my classification.
The problem is that I don't know which normalization is good in the classification study? Is it good to use FPKM? What about using DESeq2 normalization?
I would appreciate if someone could advise me, Many thanks