DESeq error that has me worried.

[tirohia]

So I'm reading in data like this:

datafile<-file.choose() # choose input file
counttable = read.table( datafile,sep=",", header=TRUE, row.names=1 )
head(counttable)

I get the design working and proceed with the following:

pairedSamples = design$libType == "paired-end"
countTable = counttable[ , pairedSamples ]
condition = design$condition[ pairedSamples ]
head(countTable)

that last line returns:

        untreated1 untreated2 untreated3 untreated4 untreated5 treated1 treated2 treated3 treated4 treated5
gene5518002       2370        804       1621       1824       3571     2920     2228     4168     1197     1981
gene5518003        624        439        710        967       1162     1076      748     1120      304      624
gene5518004        284        104        356        627        644      301      284      487       64      348
gene5518005         17         22         38         59         16       63       11       14       14       34
gene5518007          0          2          0          6          0        4        0        4        0        0
gene5518008        611        230        500        861        962      674      518     1130      201      806

The problem occurs here:

cds <- newCountDataSet(countTable, conds)

which returns:
Error in round(countData) : Non-numeric argument to mathematical function

This isn't a new error. If I do a bit of digging around the general consensus appears to be that for some reason countData is fitting the letters in the gene name and spitting the dummy.

When I'm reading my table in though I specify row.names=1 specifically so that this sort of thing shouldn't happen. If I rename all my genes to purely numeric, i.e. get rid of the word gene, everything works fine.

After reading http://seqanswers.com/forums/showthread.php?t=26598 it makes me suspect that my now numeric gene names might be being used as gene counts which would screw things up spectacularly. Given that I've removed the letters from the gene names and everything works, it implies that there's nothing wrong with the rest of my data.

Coincidently, when I run through the viginette with the Pasilla data, the everything works fine, and there are letters in the gene names.

Is anyone able to enlighten me as to what might be going wrong here?

[gringer]

Does this also happen when you add 'stringsAsFactors = FALSE' to the table read function? It sounds a bit like a data formatting error.

[tirohia]

Yup. Adding that to the table read didn't change anything.

My first thought was data formatting - given that it works fine for the pasilla viginette. After having removed all of the letters from the gene names though, it runs fine. Which makes me think there's nothing wrong with the rest of my data.

Which means for some reason that I can't fathom, it's trying to use the first column as read counts rather than genenames. Or at least, i think that's what it's trying to do.

Cheers
Ben.