Analyzing data that has previously been SCT normalized

[acunha92]

I am trying to do some differential expression analysis from scRNA-seq dataset that is publicly available: https://eqtlgen.org/sc/datasets/1m-scbloodnl-dataset.html

They mention that the data has already been QCed and SCT normalized in a matrix format. I am confused as to how to approach analyzing such a count matrix. When I CreateSeuratObject() with the matrix there is no scale.data() layer and PCA and any other downstream analysis can not be completed.

Am I expected to still run SCT() to generate the layer or is there a way to call the layer directly from the count matrix?

Thank you!

[georgia-miller]

Hi,

I've actually been analysing that same dataset! I got over this issue by moving the already normalised data into the "data" layer (removes need for normalisation as you shouldn't renormalise), then for scaling you can set scaling and centring to false so it moves into the correct scale.data layer so PCA and downstream analyses can run as normal. My code reads:

  # move counts data into the data layer (without performing normalisation)
  subset <- SetAssayData(object = subset, layer = "data", new.data = subset@assays$RNA@layers$counts, assay = "RNA")
  
  # move data to the correct scale.data layer (without actually scaling)
  all_genes <- rownames(subset)
  subset <- ScaleData(subset, features = all_genes, do.scale = FALSE, do.center = FALSE)

  # run PCA as normal
  subset <- RunPCA(subset, features = VariableFeatures(object = subset))

[acunha92]

Thank you so much! This seems to work so far. I had to resort to ScaleData() with the default variable features because of memory limitations, but I think this will be a great starting point.