classify_cells: Classify cells from multiple models

In grisslab/scAnnotatR: Pretrained learning models for cell type prediction on single cell RNA-sequencing data

Classify cells from multiple models

Description

Classify cells from multiple models

Usage

classify_cells(
  classify_obj,
  assay,
  slot = NULL,
  classifiers = NULL,
  cell_types = "all",
  chunk_size = 5000,
  path_to_models = "default",
  ignore_ambiguous_result = FALSE,
  cluster_slot = "clusters"
)

Arguments

classify_obj	the object containing cells to be classified
assay	name of assay to use in classify_object
slot	type of expression data to use in classify_object. For Seurat object, some available types are: "counts", "data" and "scale.data".
classifiers	list of classification models. The model is obtained from train_classifier function or available in current working space. Users may test the model using test_classifier before using this function. If classifiers contain classifiers for sub cell types, classifiers for parent cell type must be indicated first in order to be applied before children classifiers. If classifiers is NULL, the method will use all classifiers in database.
cell_types	list of cell types containing models to be used for classification, only applicable if the models have been saved to package.
chunk_size	size of data chunks to be predicted separately. This option is recommended for large datasets to reduce running time. Default value at 5000, because smaller datasets can be predicted rapidly.
path_to_models	path to the folder containing the list of models. As default value, the pretrained models in the package will be used. If user has trained new models, indicate the folder containing the new_models.rda file.
ignore_ambiguous_result	return all ambiguous predictions (multiple cell types) to empty When this parameter turns to TRUE, most probably predicted cell types will be ignored.
cluster_slot	name of slot in meta data containing cluster information, in case users want to have additional cluster-level prediction

Value

the input object with new slots in cells meta data New slots are: predicted_cell_type, most_probable_cell_type, slots in form of [cell_type]_p, [cell_type]_class, and clust_pred (if cluster_slot was provided).

Examples

# load small example dataset
data("tirosh_mel80_example")

# train one classifier for one cell type, for ex, B cell
# define genes to use to classify this cell type
selected_marker_genes_B = c("CD19", "MS4A1", "CD79A")

# train the classifier
set.seed(123)
classifier_b <- train_classifier(train_obj = tirosh_mel80_example,
assay = 'RNA', slot = 'counts', marker_genes = selected_marker_genes_B,
cell_type = "b cells", tag_slot = 'active.ident')

# do the same thing with other cell types, for example, T cells
selected_marker_genes_T = c("CD4", "CD8A", "CD8B")
set.seed(123)
classifier_t <- train_classifier(train_obj = tirosh_mel80_example,
assay = 'RNA', slot = 'counts', marker_genes = selected_marker_genes_T,
cell_type = "T cells", tag_slot = 'active.ident')

# create a list of classifiers
classifier_ls <- list(classifier_b, classifier_t)

# classify cells with list of classifiers
seurat.obj <- classify_cells(classify_obj = tirosh_mel80_example,
assay = 'RNA', slot = 'counts', classifiers = classifier_ls)

grisslab/scAnnotatR documentation built on March 20, 2023, 2:42 a.m.