Background choice for GO analysis of RNA-Seq

Question

[Júlia]

Hello everyone! I have a question about which Background to use for the Gene Ontology Analysis of the DEGs obtained by RNA-Seq. According to the instructions on the GO website:

" The reference list should be the list of all the genes from which your smaller analysis list was selected. For example, in a list of differentially expressed genes, the reference list should only contain genes that were detected at all in the experiment, and thus potentially could have been on a list of genes derived from the experiment."

In this context, I'm using the list of genes obtained by the exactTest() function of edgeR, which gives a list based on "pseudo - counts". Since this "pseudo-counts" are for internal use of edgeR pipeline, I don´t know if it is adequate in this case. Any thoughts?

Thank you!

[Jean-Karim Heriche]

Did you read the paper mentioned in this post?

Answer 1

[jared.andrews07]

Run filterByExpr prior to testing and roll with the remaining genes as your background.

[Júlia]

Hi! We actually use another filtration:

y <- DGEList(counts=count, group=group) 
#Normalization
y <- calcNormFactors(y) 
#Filtering
keep <- rowSums(cpm(y)>1) >= 3 
y <- y[keep, , keep.lib.sizes=FALSE]
# Calculate dispersions 
y <- estimateCommonDisp(y)
y <- estimateTrendedDisp(y)
y <- estimateTagwiseDisp(y)
# Testing for DE genes 
et <- exactTest(y, pair=c("control", "case"))

This approach works as well? We have four samples in case group and six samples in controls, with only one factor. Thanks in advance!

[jared.andrews07]

Well, you're free to take whatever approach you like. Just use whatever passes your filtering criteria as the background.