Hello,
I have a question about DESeq2 designs and implementation. first of all, Michael Love you did a wonderful job with the package and vignette - thank you !.
My RNA-Seq experiment has 24 samples: 4 models, 2 treatments (drug-treated or control), each has 3 biological replicates. Two of the models "A", "B" responded while "C","D" did not. I'm trying to figure out the right design to check what genes are differentially expressed in responders vs Non-responders after treatment.
After reading many posts and the DESeq2 manual I thought this is similar to Group-specific condition effects, individuals nested within groups section.
Q1: am I thinking about this right or there is a simpler design? mostly interested in what if the drug effect in R vs NR
Q2 (optional): if this is the way to go I can't find the right contrast to get... code below
response <- factor(rep(c("Response","NoResponse"),each=12))
Model <- factor(rep(c("A","B","C","D"),each=6))
treatment <- factor(rep(rep(c("Drug","NoDrug"),each=3),4))
# also adding ind as described in DESeq2 manual
ind.nested <- factor(rep(rep(1:2,each=6),2))
coldata1<- data.frame(response,Model,treatment,ind.nested)
m1 <-model.matrix(~ response + response:ind.nested + response:treatment, coldata1)
# object construction - simple design (later replaced)
dds<-DESeqDataSetFromMatrix(countData=exprs(eset1),colData=pData(eset1),design = ~ Model + treatment)
# DE
deg<- DESeq(dds,full = m1,betaPrior = F)
# I don't understand which interaction/contrast do I need to pull?
resultsNames(deg)
resultsNames(deg)
[1] "Intercept" "responseResponse" "responseNoResponse.ind.nested2"
[4] "responseResponse.ind.nested2" "responseNoResponse.treatmentNoDrug" "responseResponse.treatmentNoDrug"
I think I'm looking for contrast=list("responseNoResponse.treatmentDrug","responseResponse.treatmentDrug") but I dont have it.
I appreciate your help,
Elad
