I’m a PhD student struggling to get my head around what I do to normalise my circRNA data (have tried to look at all of the related posts on various sites)
I am using DESEQ2 for differential analysis and have my gene matrix counts for the linear genes:
[1] dds <- DESeqDataSetFromTximport(txi, design_matrix, ~pair + cond)
[2] dds<- estimateSizeFactors(dds)
[3] dds<- DESeq(dds,parallel=TRUE)
[4] res.gene<- results(dds)
I now have a separate raw count matrix for the backsplice junction reads for my circRNAs.
To normalise would you suggest either of the following or something different?
Before line 2 (above) concatenate the circ raw counts matrix to the linear raw counts matrix for all samples and then proceed
Or
Somehow extract the SizeFactors and apply them to the circRNA count matrix? (If so how would I extract the size factors?)
Or Something different?
You advice would be GREATLY appreciated!
Zoe.
Hi Michael, I have data from a group of patients measured at two time points (paired analysis). I want to identify circs that are differentially expressed between the two time points i.e I want to find circRNAs in which expression at time point B is larger/smaller than time A controlling for library size.
My understanding is that when analysing circs you can't normalise for library size on the total circRNAs alone.
Ok so I've got the DE between time points aspect. I just don't know what's the state of the art for normalization. So instead of looking across circRNA, you want to estimate size factors on the standard counts matrix?
That's easily done:
Thanks Michael, I'm getting this error though? Error in (function (classes, fdef, mtable) : unable to find an inherited method for function ‘sizeFactors<-’ for signature ‘"DESeqDataSet", "NULL"’
Just to be clear, you'll need to define the
dds1
anddds2
above to be relevant to your dataset.The code example is if you wanted to apply the size factors from one dataset to another. So I'm just showing how DESeq2 can be used. I don't happen to know the optimal normalization for circRNA, so you'll have to ask others for best practices. Here I provide software support.
Hi Michael,
I have used tximport to get my deseq object. Am I still able to extract the library size using sizeFactors()???
Not exactly, this has come up on the support site before, so see if you can find previous threads. Short answer is that you can get normalizationFactors() instead.