Converting SingleCellExperiment objects into Seurat objects

[mass-a]

Hello,

I am having trouble converting SingleCellExperiment objects from the scRNAseq bioconductor collection using as.Seurat. Here's a reproducible example:

BiocManager::install('scRNAseq')
library(scRNAseq)

sce <- ZilionisLungData('mouse')
seurat <- as.Seurat(sce, counts = "counts", data=NULL)

Error in embeddings[i, j, drop = drop, ...] : subscript out of bounds

Several other datasets in the collection give similar errors. Any idea what's going on here? can I e.g. ask as.Seurat to only convert counts and metadata?

Thanks in advance,
-massimo

[alikhuseynov]

Hi,
..same issue as well.
it only works when removing embeddings (maybe something is wrong with them in that sce object)

library(scRNAseq)
sce <- ZilionisLungData('mouse')
reducedDim(sce) <- NULL # remove dr
object <- as.Seurat(sce, counts = "counts", data = NULL)
An object of class Seurat 
28205 features across 17549 samples within 1 assay 
Active assay: originalexp (28205 features, 0 variable features)

# package versions
lapply(seq(2), function(i) package.version(c("SeuratObject", "Seurat")[i])) %>% unlist
1] "4.1.3"      "4.2.1.9000"

[timoast]

The problem with this dataset is many of the cell names are duplicated:

library(scRNAseq)
sce <- ZilionisLungData('mouse')
length(unique(colnames(sce)))
[1] 2711
length(colnames(sce))
[1] 17549

This is unusual, and we require that within a Seurat object a cell name is unique to one cell. The issue here is that we are automatically renaming cells with conflicting names, but that renaming is not being performed on the dimension reduction object