Feature Request: Change orig.ident to numeric if possible before factoring

[jdrnevich]

Motivation

There is a common situation for us that leads to incorrect factor leveling. We routinely read in aggregated outputs from cellranger, which adds on -1, -2, etc. endings for each sample. CreateSeuratObject() conveniently has names.delim and names.field arguments we can use to get the endings as the orig.ident. However, when there are 10+ samples, the factor leveling order is incorrect because they were character data when factored:

> aggr_data <- Read10X(data.dir = "results/cellranger/Aggregated/outs/count/filtered_feature_bc_matrix/")
> allcells <- CreateSeuratObject(counts = aggr_data, 
+                                names.delim = "-",
+                                names.field = 2)
> levels(allcells$orig.ident)
 [1] "1"  "10" "11" "12" "13" "14" "15" "16" "17" "18" "2"  "3"  "4"  "5"  "6"  "7" 
[17] "8"  "9" 

Feature Description

This causes problems when trying to switch to sample IDs or other metadata that is in the aggregation.csv. Looks like the relevant code in CreateSeuratObject in https://github.com/satijalab/seurat-object/blob/main/R/seurat.R is on lines 1349:1354. Instead of

# Create identity classes
  idents <- factor(x = unlist(x = lapply(
    X = colnames(x = counts),
    FUN = ExtractField,
    field = names.field,
    delim = names.delim
  )))

You could check to see if they can be converted to numeric before factoring with something like:

# Create identity classes
  idents <-x = unlist(x = lapply(
    X = colnames(x = counts),
    FUN = ExtractField,
    field = names.field,
    delim = names.delim
  ))
# Change to numeric if possible
if(!any(is.na(suppressWarnings(as.numeric(idents )))) {
    idents <- as.numeric(idents)
}
idents <- factor(idents)

I'm not sure how this would work with the next checks on idents so I didn't submit it as a pull request

Alternatives

No response