edgeR: no residual df: Cannot estimate dispersion

Question

[dec986]

Hello,

I have a data set looking like this:

Gene Symbol a1  a2  b1  b2
Prss12  475 347 342 367
2010008C14Rik   3   0   0   0
Gm10819 216 129 164 126
Gm2245  25  23  14  28

and I am loading into EdgeR with the following script:

#!/usr/bin/R
counts = read.table("gene_count.tsv",sep="\t",row.names=1,header=T)
group <- c("a1","a2","b1","b2")
gene_or_transcript <- "genes_"#for filenames so I know what I'm reading
countS = counts[,1:dim(counts)[2]]
library(edgeR)
#now we'll do the DGEList command
y = DGEList(counts=counts[,1:dim(counts)[2]] ,gene=row.names(counts))#counts=counts[,2:dim(counts)[2]] removes $Gene.Symbol, i.e. 1st column
o<-order(rowSums(y$counts),decreasing=TRUE)#remove duplicates
y<-y[o,]
d<-duplicated(y$genes)
y<-y[!d,]
y$samples$lib.size <- colSums(y$counts)
rownames(y$counts) <- t(y$gene)# exchange ID
y<-calcNormFactors(y)
data.frame(Sample=colnames(y),group)
design<- model.matrix(~0+group)
#
y<-estimateGLMCommonDisp(y,design, method = 'deviance')
y<-estimateGLMTrendedDisp(y,design)
y<-estimateGLMTagwiseDisp(y,design)

but this produces an error

"Error in return(NA, ntags) : multi-argument returns are not permitted
In addition: Warning messages:
1: In estimateGLMCommonDisp.default(y = y$counts, design = design,  :
  No residual df: setting dispersion to NA
2: In estimateGLMTrendedDisp.default(y = y$counts, design = design,  :
  No residual df: cannot estimate dispersion"

why is this error produced? how can I fix it? I don't see anything in the EdgeR manual about it.

thanks, -DEC

Answer 1

[Devon Ryan]

You only have 4 samples, you can't then have 4 groups. I presume you meant group <- c("a", "a", "b", "b").

[dec986]

thank you very much Devon!