[BioC] Questions about GOseq after DESeq2

[amandine.fournier at chu-lyon.fr]

Dear Bioconductor users and developers,

I have 2 questions regarding a gene ontology analysis of RNA-Seq data.

I first used the DESeq2 package to perform the differential analysis of my data and it resulted in a list of about 500 DEG. Now I would like to perform a gene ontology analysis on this dataset with the GOseq package.

According to the GOseq user manual, this package requires 2 pieces of information : the differentially expressed genes (= my list of 500 genes) and the measured genes. These are defined as "all genes for which RNA-seq data was gathered for your experiment".

As you known, DESeq2 filters out low counting genes and outliers. So I wonder if I should consider :
- all sequenced genes including low expressed genes and outliers, 
- or the sequenced genes whithout outliers (but including low expressed genes)
- or only remaining genes after low count filtering (final assayed genes) ?

I don't know which step filters outliers out. It seems that they are taking into account to estimate library sizes, but I am not sure if they are filtered before or after estimating dispersion values.
Low expressed genes are filtered out at the end of the analysis, juste before calculating FDR, so they are taking into account for calculating library sizes, estimating dispersion values, fitting the GLM model and testing, but not in the final FDR.
In brief, I wonder if outliers and low expressed genes can be considered as "assayed" in the DESeq2 analysis.

Do you have an opinion about this ? Is there a commonly accepted / advised methodology for that ?

The second question is : should I analyse over-expressed genes and under-expressed genes together or separately ?

Many thanks in advance.
Best regards,
Amandine

-----
Amandine Fournier
Lyon Neuroscience Research Center
& Lyon Civil Hospital (France)