I have a project on Rice and my conditions (paired-end) unfortunately have no replicate.
I used tophat2 and after conversion to sam file is used HTSeq and I merged HTSeq output.
And today the bigest problem emereged, How can i handle DeSeq without replicate you can see my following codes:
pasillaCountTable <- read.table("MyCountTable.txt") pasillaDesign = data.frame(row.names = colnames( pasillaCountTable ),condition = c( "C1","S1"),libType = c( "paired-end", "paired-end" )) pairedSamples = pasillaDesign$libType == "paired-end" countTable = pasillaCountTable[ , pairedSamples ] condition = pasillaDesign$condition[ pairedSamples ] cds = newCountDataSet( countTable, condition ) cds = estimateSizeFactors( cds )
and main part was, is it fine or not?
cds = estimateDispersions( cds, method="blind", sharingMode="fit-only" ,fitType="local")
and rest of the codes:
res_C1S1 = nbinomTest( cds, "C1", "S1" ) write.table(res_C1S1,"C1_S1.txt",sep= "\t")
Thanks a lot.
If you want to do analysis of differentially expressed genes, that just won't be possible without replicates. If you don't know what's the variance within the samples from the same group (either condition or control), you cannot possibly tell if the difference you see between different groups (condition vs control) is significant or not. You may try to take a look at dispersion values from the very similar experiment if such data is available.