Is DeSeq2 or EdgeR better for differential expression of QuantSeq data?

Question

[james.zhang20]

Hi everyone,

I am starting to analyse an RNA-Seq dataset, which was prepared by QuantSeq (3' sequencing), which generates one fragment per transcript, such that the read count is directly proportional to the number of transcripts (https://www.nature.com/articles/nmeth.f.376). I have four different conditions and three replicates between them, and ideally would like to carry out a differential expression analysis. I found this source (https://hbctraining.github.io/DGE_workshop/lessons/02_DGE_count_normalization.html) which suggests that DeSeq2's median of ratios would not generally be suitable for this, as it does not normalise for gene length, but as QuantSeq data does not need to be normalised to gene length anyway, would it be better to use DeSeq2, as EdgeR includes a superfluous gene length normalisation?

Thanks!

Answer 1

[ATpoint]

Both are fine. There is no difference for gene level ‚between-sample’ differential analysis between these end-tagged (like QuantSeq) or full length RNA-seq methods. Normalization if both tools is very similar and suitable, gene length plays no role here, nor is it done by any if the common RNA-seq tools anyway, even for full length protocols. edgeR does not normalize for any sort of gene length, you’re misquoting or misunderstanding something here. Use any of the two tools you like more, both are fine.

Answer 2

[swbarnes2]

There is a section right in the tximport vignette about how to deal with 3' counting methods, and how to not introduce a gene length offset.

https://bioconductor.org/packages/devel/bioc/vignettes/tximport/inst/doc/tximport.html#3%E2%80%99_tagged_RNA-seq