Label Clusters in Seurat

Question

[planetearth]

I have 2 plots, a control and stimulated group of cells.

After clustering, the cluster labels are 0, 1, 2.

EDIT How can I know what cell types are in each cluster? The known cell type names are in the rows of my data matrix, but how do I search for their names in the cluster?

>DimPlot(stem.combined, reduction = "umap", split.by = "control_subset")

Thanks

Answer 1

[jared.andrews07]

This should do it.

# Save old identity classes (the cluster labels) for reference.
stem.combined[["old.ident"]] <- Idents(object = stem.combined)

# Rename classes.
stem.combined <- RenameIdents(object = stem.combined, `0` = "your cell type", `1` = "your other cell type", `2` = "your last cell type")

[planetearth]

Thanks, but how would I know which cluster is which cell type? Is there a way to find the identity of my clusters?

[jared.andrews07]

Oh, your original question made it sound like you already knew the types. You can look at the markers that best define each group via FindAllMarkers and use your expert knowledge of which cell types you expect to manually assign them. Or look at marker genes for various cell types that you expect just using FeaturePlot.

For an inference-based approach that uses reference expression profiles from dozens of different cell types, you can check out the R package SingleR. It is currently being prepped for Bioconductor submission and works quite well, particularly for immune cells. If you choose to use it, it's important to note the README is out of date. It should be installed via devtools::install_github("LTLA/SingleR").

Answer 2

[AB]

Another option is to use Garnett by the Trapnell lab, the same people who gave us monocle. It uses machine learning to train classifiers for a certain cell type and apply those classifiers to a single cell dataset