cell type annotation

[igordot]

Is it possible to add the cell types to the metadata?

[federicomarini]

Hi @igordot!
Since this information was not provided in the original data release, I did not include that back then in the original package submission.
As of now, the SingleR method is available in Bioconductor and can become a nice standard solution to address this.

I'll note it down to be revisited at some point, thanks for the suggestion! (ideally, an expansion of the vignette could be a way to cover this)

[igordot]

Thanks for considering it! I assumed that type of information would be included. I mostly just wanted to makes sure I wasn't missing anything.

[federicomarini]

Just gave a quick trial run, and this seems to work.

Starting on the code from the vignette (say, sce is your bone marrow dataset, subset to keep it quick in the first run):

library(iSEE)
library(SingleR)
se_hpca <- HumanPrimaryCellAtlasData()
common <- intersect(rownames(sce), rownames(se_hpca))
sce_anno <- sce[common,]
se_hpca <- se_hpca[common,]
sce_anno
library(scater)
sce_anno <- logNormCounts(sce_anno)
pred <- SingleR(test=sce_anno,
                ref=se_hpca,
                labels=se_hpca$label.main,assay.type.ref = "logcounts")
table(pred$labels)
sce_anno$labels_sr <- pred$labels
plotTSNE(sce_anno, colour_by="labels_sr",text_by = "labels_sr")

(using current Rdevel setup)