RNAseq count which strand to use

Question

[leisureh]

naive question, but do not find a clear answer to solve this,

downloaded a raw count file from geo, which looks like this,

gene   unstranded   fwd   rev
ENSMUSG00000102693  0   0   0
ENSMUSG00000064842  0   0   0
ENSMUSG00000051951  63  38  25

How could I use this file to generate the count table? or I can just simply use the count of the unstranded?

Thanks

Answer 1

[lieven.sterck]

In theory you should first check the metadata of that experiment and check if they had stranded or un-stranded data .

If that is not available for whatever reason, I would personally go with the unstranded data indeed (which was the common mode before we switched to doing stranded seq by default) .

If you want to go for the stranded one, you will have to look up the strand of each gene and take the count that is reported for that strand.

Answer 2

[swbarnes2]

You want to look at more rows of data to be sure (obviously the genes with 0 counts tell you nothing), but that pattern of the left, unstranded column having about as many reads as the forward and reverse columns together suggests that your data is unstranded, so that is the correct column to use. If your data were stranded, the unstranded count would be just about what either the forward or reverse columns would be, and the other direction would be mostly empty. (Like 63,0,63 or 63,63,0)

[Istvan Albert]

just to add to the confusion, if the data were paired-end, then the two counts would also be very close regardless of whether the sequencing was stranded or not

[swbarnes2]

True. I forgot, because my lab doesn't do paired end for gene counting. If you had a bam you could sort that out, but if you don't you have to just go with unstranded counts.

[chrisgr]

My data has both the sum and 'either one' like you explained. Then what collumn should I take? :)