Add GEX data without normalizing

[mitchTwoTimes]

Is it possible to addGeneExpressionMatrix without normalizing? I was wondering if I could maintain the raw counts within ArchR and save normalization for when I export to Seurat. I tried passing NULL to scaleTo and received the following

Error in data.frame(scaleTo = scaleTo, exclude = excludeChr, stringsAsFactors = FALSE) : 
  arguments imply differing number of rows: 0, 2

Sorry I can't provide more - my RStudio session crashes before I can get the full traceback().

[rcorces]

This isnt currently possible. And it would probably muck around with some of the base assumptions of other ArchR functions which assume that the RNA data is normalized. In release_1.0.2, scaleTo is not allowed to be NULL:

ArchR/R/MatrixGeneExpression.R

Line 44 in 9b3e72e

.validInput(input = scaleTo, name = "scaleTo", valid = c("numeric"))

But that .validInput() call was added relatively recently. I think it should be relatively simple to get the un-normalized counts back out by reversing the math done here for normalization:

ArchR/R/HiddenUtils.R

Line 178 in 968e442

mat@x <- scaleTo * mat@x / rep.int(colSm, Matrix::diff(mat@p))

But I'm not sure that helps you very much.

[mitchTwoTimes]

Ok thanks for sharing that hack. Is there no log transformation that occurs?

[rcorces]

yes sorry there is also a log2(count+1) transform that occurs here:

ArchR/R/ArrowWrite.R

Line 206 in 968e442

mat@x <- log2(mat@x + 1) #log-normalize