I'd like to manually add a vector of transcript sizes to my DDS object. I tried setting:
dds@assays$data$avgTxLength=data.frame(row.names=txOrder,"avgTxLength"=tlen[txOrder,])
Where txOrder is the row names from the dds data and tlen is the (out of order) transcript list for those row names. The error I get when I try using fpkm() is "invalid 'dimnames' given for data frame".
How do I add this vector so I can output fpkm?
hi,
You say vector, which sounds like you just have a single number per gene. Is that correct? In that case, if you take a look at ?fpkm, the thing to do is just:
mcols(dds)$basepairs <- x
Where x is your vector of gene lengths, lined up with rows of dds.
Your DDS is a subclass of a summarizedexperiment. rowData(DDS) Should return you a DataFrame of gene info. So, you should be able to do something like rowData(DDS)$txlength <- xxx
Where xxx is a properly ordered vector of gene lengths
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
Nope, neither worked.
I'm working at the transcript level and I have the length for each transcript (calculated with a simple script from the gtf). So tlen is a data frame with transcript_id's as row names and one column of data labeled "Len". txOrder is a character array with the transcript_id values used in my dds object (after filtering) in the correct order.
Replacing "xxx" or "x" in the above examples with tlen[txOrder,] (which produces a numeric vector) still gives the "dimnames" error when I try fpkm().
Can you make a small reproducible example, so I can take a look? Ideally with just simulated data, e.g. makeExampleDESeqDataSet.
In creating a subset example object, I can now get the mcols(dds)$basepairs<-x method to work. So now I'll go back and re-create my original dds object--perhaps I screwed it up with my other trial methods. But thanks--this seems to have solved my problem!
For bp gene lengths, should I use the original CDS length or the effective CDS length (CDS length - read length)? Thanks!
I don't have any preference here. I tend to use methods like Salmon for calculating abundance, rather than FPKM from counts.