Removing replicate not clustering and group with replicate Vs without -edgeR rnaseq analysis

Question

[veronica]

I am working with bacteria samples - in 3 groups that include the control, Treatment A, and Treatment B. From the PCA I find that the replicates are far apart. So I have removed the treatment A replicate r2 and treatment B r2 replicate and used edgeR (qCML) n=2 Vs n=1 analysis. I know rnaseq without sufficient replicates is not of any value But I want a suggestion if I can still use the data only to select those genes with log2fold change >2 and <-2?

or use GFOLD ( or gfold diff -s1 sample1,sample2 -s2 sample3 -suf .read_cnt -o 12VS3.diff) and present only the generalized log fold for selected genes. Any advice?

group<-c(1,1,2,3)
y <- DGEList(counts=x, group=group)
y <- estimateDisp(y)
et <- exactTest(y,pair=c(1:2)
 topTags(et)