condition of use about RNA-Seq normalisation

[ickou]

Hi,

I use 3 tools to analyse RNA-Seq data :
- edgeR
- DESeq
- NOISeq

Each tool offers several normalisation :
- TMM (Trimmed Mean of M)
- RLE (relative log expression) = by default in DESeq ?
- quantile (edgeR) = Upper Quartile (NOISeq) ? Counts are divided by the third quartile of
counts for transcripts with at least one read ?

I know TMM method assume that the majority of gene are not differentially expressed...
In my case, I analyse data with a lot of gene that are differentially expressed.
Can I use RLE normalisation with this data ? TMM method is not appropriate.
I think Upper-Quartile is OK for my case.

Thank you for your help

Emeric