scROSHI {scROSHI}

Robust Supervised Hierarchical Identification of Single Cells

Description

Identifying cell types based on expression profiles is a pillar of single cell analysis.

scROSHI identifies cell types based on expression profiles of single cell analysis by utilizing previously obtained cell type specific gene sets. It takes into account the hierarchical nature of cell type relationship and does not require training or annotated data.

Usage

scROSHI(
  sce_data,
  celltype_lists,
  type_config,
  count_data = "normcounts",
  gene_symbol = "SYMBOL",
  cell_scores = FALSE,
  min_genes = 5,
  min_var = 1.5,
  n_top_genes = 2000,
  n_nn = 5,
  thresh_unknown = 0.05,
  thresh_uncert = 0.1,
  thresh_uncert_second = 0.8,
  verbose = 0,
  output = "sce"
)

Arguments

sce_data

A SingleCellExperiment object containing the expression profiles of the single cell analysis.

celltype_lists

Marker gene list for all cell types. It can be provided as a list of genes with cell types as names or as a path to a file containing the marker genes. Supported file formats are .gmt or .gmx files.

type_config

Config file to define major cell types and hierarchical subtypes. It should be provided as a two-column data.frame where the first column are the major cell types and the second column are the subtypes. If several subtypes exists, they should be separated by comma.

count_data

Assay name in the SingleCellExperiment object containing the count data.

gene_symbol

Variable name in the row data of the SingleCellExperiment object containing the gene names.

cell_scores

Boolean value determining if the scores should be saved.

min_genes

scROSHI filters out non-unique genes as long as more than min_genes are left. If there is a cell type that has less than min_genes genes, it will be replaced with the cell type list BEFORE filtering for unique genes (default 5).

min_var

Minimum variance for highly variable genes (default 1.5).

n_top_genes

Maximum number of highly variable genes (default 2000).

n_nn

Number of nearest neighbors for umap for assignment of cell types (default 5).

thresh_unknown

If none of the probabilities is above this threshold, the cell type label is assigned to the class unknown (default 0.05).

thresh_uncert

If the ratio between the largest and the second largest probability is below this threshold, the cell type label is assigned to the class uncertain for the major cell types (default 0.1).

thresh_uncert_second

If the ratio between the largest and the second largest probability is below this threshold, the cell type label is assigned to the class uncertain for the subtypes (default 0.8).

verbose

Level of verbosity. Zero means silent, one makes a verbose output.

output

Defines the output. sce: The output is a SingleCellExperiment object with the cell types appended to the meta data. df: The output id a data.frame with two columns. The first column contains the barcode of the cell and the second column contains the cell type labels.

Examples

data("test_sce_data")
data("config")
data("marker_list")

results <- scROSHI(sce_data = test_sce_data,
                   celltype_lists = marker_list,
                   type_config = config)
table(results$celltype_final)