Entering edit mode
Hi,
We have RNAseq data for ~120 samples.
We are considering of using "Trinity Differential Expression" protocol described by Brian Haas in https://github.com/trinityrnaseq/trinityrnaseq/wiki/Trinity-Differential-Expression
How well does it work without the RSEM abundance of estimation?
We were thinking of feeding normalized gene counts (e.g., by TMM logCPM) into the Trinity protocol.
Thank you very much for your help,
Arik
Not being the developer, but I would recommend to do these kind of analysis manually rather than in a closed end-to-end pipeline where you essentially have no control over any parameters or details. You are probably interested in differential gene rather than transcript-level expression? I'd start from raw gene-level counts, e.g. obtained via
tximport
, not normalized counts as these are not compatible with most common DE tools as explicitely mentioned in their vignettes. Be sure to properly perform QC to detect potential batch effects and outlier samples.Thank you.
Regarding: "I'd start from raw gene-level counts, e.g. obtained via tximport" Did you mean without RSEM fix for abundance?
The protocol is recommending to use logcpmTMM normalization, but they do it on rsem counts not. I was wondering how important is that the input will be rsem. We have allready did the DE analysis with dseq2.