13 months ago by
CRUK, Cambridge, UK
I am currently working on giving users much more control over normalization in
DiffBind. In the meantime, there is a trick you can do you avoid normalization when using
DESeq2 as the analysis method.
dba.count(), you can change the individual library sizes to all be the same. Then if you run
bFullLibrarySize=TRUE (default) the normalization facts will all be equal to 1. You probably want to run
bSubControl=FALSE to avoid having the control reds subtracted if do not want normalization of any kind.
You can set the library size values as follows:
> myDBA <- dba.count(myDBA, ...)
> myDBA$class[8,] <- 1e+07
> myDBA <- dba.analyze(myDBA, method=DBA_DESEQ2, bFullLibrarySize=TRUE, bSubControl=FALSE)
Also, there is a "backdoor" in
DiffBind to allow you to provide your own normalized read counts -- if that is of interest, I can show the code required to use it.