DESeq2 single cell pseudobulk

Question

[fluentin44]

Hi,

So im slightly confused about which counts data from a Seurat object to use for further interrogation. My understanding is:

• 'counts' slot of 'RNA' assay = Raw counts
• 'data' slot of 'RNA' assay = Normalised counts
• 'scale.data' slot of 'RNA' assay = Normalised and scaled counts

When conducting pseudobulk analysis with (aggregated) single cell data its been reccomended to use the raw counts from the 'RNA' assay - does this mean all of the variables you may have regressed out during the QC, i.e. cell cycle difference, culture batch, sex would need to be added to the model?

Thanks, Matt

Answer 1

[ATpoint]

The raw counts will be summed to pseudobulks, so any additional factor or covariate would need to be added again. Still, since a bulk is a sum over the cells something like cell cycle does not necessarily manifests the same way as it would on single-cell level. I'd suggest doing the usual QC such as PCA to check whether you can get some evidence that mentioned covariates are necessary to adjust for.

[fluentin44]

Much appreciated, thanks.