GAGE: no essential member genes for some DE pathways
1
0
Entering edit mode
@aurelienicolas-13013
Last seen 5.4 years ago

Hello,

I have meta-transcriptomics data (RNAseq) and I used GAGE (v2.26.0) to find KEGG ortholog pathways differentially expressed. I have lists of KEGG pathways perturbed (in 1 and 2 directions)  and I want to extract significantly DE gene list (in the same way as DAVID output). So, I use essGene function but sometimes I have no genes for a pathway with p.val < 0.1. I use same options in gage and essGene function as you can see in this extract of my code :

RESgage.kegg.genus.both <- gage(exprs=ExpressionData.ko.genus, gsets=kegg.sets.ko, ref=lf, samp=lt, compare = "paired", same.dir = FALSE )
output.bidir=RESgage.kegg.genus.both.select[RESgage.kegg.genus.both.select\$p.val < 0.1, 1:5]
listPathway=rownames(output.bidir)
geneSet<-vector(mode = "character",length = length(listPathway))
for (j in 1:length(listPathway)){
termselec=listPathway[j]
gs=unique(unlist(kegg.sets.ko[termselec]))
essData = essGene(gs, ExpressionData.ko.genus, ref=lf, samp=lt,  compare= "paired")
essDataList=rownames(essData)
geneSet[j] <- toString(essDataList)
}
output.bidir=cbind(output.bidir, geneSet)


If I set use.fold=FALSE in essGene, I obtain genes but I want DE genes considered by gage function. Can you help me ?

I don't understand how gage function can considere a pathway perturbed without gene significantly changed.

gage • 771 views
0
Entering edit mode
Luo Weijun ★ 1.5k
@luo-weijun-1783
Last seen 4 weeks ago
United States
Selecting pathways (or gnee sets) based on p-val instead of q-val is not good when there are multiple pathways being tested. Criterion p.val<0.1 makes things even worse. In other words, you select pathways that are not really significant. That’s likely the reason why you don’t see differentially expressed genes in the pathways. In addition, methods like GAGE or GSEA may select pathway with no individual genes significantly perturbed, but including many genes with small yet coordinated changes in 1 (or 2) direction(s). you may check the GAGE paper for details: https://bmcbioinformatics.biomedcentral.com/articles/10.1186/1471-2105-10-161