Automatic cell-type annotation

scPlant have implemented several different solutions for automatic cell-type annotation that are based on popular tools, including SingleR, Garnett, scCATCH, Celaref and CellAssign.

library(scPlant)

Reference based tools

SingleR, Celaref and CellFunTopic need reference data to perform automatic cell-type annotation.

Please see the following examples. Take seuratObj_ref as the reference scRNA-seq data, and take SeuratObj as the query un-annotated scRNA-seq data.

Using SingleR

seuratObj_ref$label <- seuratObj_ref$cell_type
SCE_ref <- Seurat::as.SingleCellExperiment(seuratObj_ref)
SeuratObj <- AutoAnnotate_SingleR(SeuratObj, SCE_ref, ref_type = "single-cell")
# The predicted cell type label is stored in SeuratObj$predicted_label.

Using Celaref

cellinfo_ref <- seuratObj_ref$cell_type %>% tibble::enframe(name = "CellId", value = "Cluster") %>% as.data.frame
result <- AutoAnnotate_Celaref(SeuratObj, counts_ref = seuratObj_ref@assays$RNA@data, 
                               cellinfo_ref = cellinfo_ref, plot = TRUE)

Using CellFunTopic

We suggest performing GSEA analysis on the reference dataset before transferring annotation to query dataset using CellFunTopic.

seuratObj_ref <- RunGSEA_plant(seuratObj_ref, by = 'GO', GeneIDtype = 'TAIR')
result <- CellFunTopic::predictFun(
  query_SeuratObj =  SeuratObj,
  seuratObj_ref,
  group_by = "cell_type",
  cluster_by = "seurat_clusters",
  species = "Arabidopsis thaliana"
)

Marker based tools

scCATCH, Garnett and CellAssign need marker genes to perform automatic cell-type annotation.

To run the following codes, a data frame with two columns (celltype, gene) is needed, containing the markers of each cell type. It is named markerDF here.

Using scCATCH

First, we need to build a data frame of marker genes with the required format.

marker_custom <- get_marker_scCATCH(markerDF, species = 'Arabidopsis thaliana', tissue = 'root')

Then, you can perform automatic annotation with scCATCH.

result <- AutoAnnotate_scCATCH(SeuratObj, marker_custom)

Using Garnett

You can build a marker file by referring to the document, or just run the following code.

get_marker_file(markerDF, SeuratObj, file = "markers.txt")
# A marker file named markers.txt will be generated in your working directory.

Then, you can perform automatic annotation with Garnett.

SeuratObj <- AutoAnnotate_Garnett(SeuratObj, marker_file_path = "markers.txt")
# The predicted cell type label is stored in SeuratObj$predicted_label.

You may run into some errors prompting format problem of the marker file while running Garnett. If so, we recommend checking the format of the marker file using check_markers function and modify the format beforehand, which will reduce useless effort and save time.

Using CellAssign

SeuratObj <- AutoAnnotate_CellAssign(SeuratObj, markerDF)
# The predicted cell type label is stored in SeuratObj$predicted_label.

2023-06-17