Run CellTypist for automated cell type annotation

Usage

sn_run_celltypist(
  x,
  celltypist = NULL,
  model = "Immune_All_Low.pkl",
  outdir = NULL,
  prefix = NULL,
  mode = c("best_match", "prob_match"),
  p_thres = 0.5,
  majority_voting = TRUE,
  over_clustering = "auto",
  min_prop = 0,
  transpose_input = TRUE,
  gene_file = NULL,
  cell_file = NULL,
  assay = "RNA",
  layer = "counts",
  xlsx = FALSE,
  plot_results = FALSE,
  quiet = FALSE
)

Arguments

x: A Seurat object or a path to a count matrix / AnnData file that CellTypist can consume.
celltypist: Path to the celltypist binary. Defaults to "/opt/mambaforge/envs/scverse/bin/celltypist".
model: Model used for predictions. Defaults to "Immune_All_Low.pkl".
outdir: Directory to store the output files. If NULL, use a temporary directory.
prefix: Prefix for the output files. By default, use the model name plus a dot.
mode: Choose the cell type with the largest score/probability ("best_match") or enable multi-label classification ("prob_match").
p_thres: Probability threshold for the multi-label classification. Ignored if mode = "best_match".
majority_voting: Logical. Whether to refine labels using majority voting after over-clustering.
over_clustering: Input file or a string key specifying an existing metadata column in the AnnData object, or "auto".
min_prop: For the dominant cell type within a subcluster, the minimum proportion of cells required to name the subcluster by this cell type.
transpose_input: Logical. If TRUE, add the --transpose-input argument when calling celltypist.
gene_file: If the provided input is in the mtx format, path to the file storing gene information. Otherwise ignored.
cell_file: If the provided input is in the mtx format, path to the file storing cell information. Otherwise ignored.
assay: Assay used when exporting Seurat counts to CellTypist. Defaults to "RNA".
layer: Layer used as the input count matrix. Defaults to "counts".
xlsx: Logical. If TRUE, merge output tables into a single Excel (.xlsx). Defaults to FALSE.
plot_results: Logical. If TRUE, plot the prediction results. Defaults to FALSE.
quiet: Logical. If TRUE, hide the banner and config info from celltypist. Defaults to FALSE.

Value

A Seurat object with three new columns in its metadata: _predicted_labels, _over_clustering, _majority_voting.

Examples

if (FALSE) { # \dontrun{
data("pbmc_small", package = "Shennong")
pbmc <- sn_run_cluster(pbmc, normalization_method = "seurat", verbose = FALSE)
pbmc <- sn_run_celltypist(pbmc, model = "Immune_All_Low.pkl")
head(colnames(pbmc[[]]))
} # }