Differential expression using DGEseq

[ckant]

First of all I am very new in this field. .
I am having RPKM values of a set of genes, in 4 different samples. I want to use DEGseq for the differential expression analysis. I would like to know, how appropriate it will be, using DEGseq for RPKM values.

[dpryan]

Firstly, don't use DEGseq. DESeq2, edgeR, or limma (see the voom() function) would be more appropriate choices.

Secondly, the ease with which one can convert between RPKM values and raw counts depends upon how the RPKM values were derived. In general, it's easier to just run featureCounts or htseq-count on the alignments and just be done with it

[sphil]

Firstly, don't use DEGseq. DESeq2, edgeR, or limma (see the voom() function) would be more appropriate choices.

In general I would say you are right not using DEGSeq but has it lost it's permit to live / to be used at all? Can't there be a case where DEGSeq seems to model the data more accuaretly than DESeq2 or edgeR. Especially since both DESeq2 and edgeR are using NB Models and DEGSeq uses poisson?

[dpryan]

It's really a question of whether the Poisson model is ever more accurate than the NB model for biologically realistic RNAseq data. I'm sure there are circumstances where that's the case, but I suspect that they're exceedingly rare. If someone is asking about converting from RPKMs to counts, then they won't/shouldn't be working on one of these rare cases

[sphil]

That's true! I general I'd say go with deseq2 / edgeR and there is litte that can go wrong.

[ckant]

Thanks

Thanks a lot for your nice suggestions, I am now using DESeq2, and its working !!!