Skip to contents

Run scDblFinder doublet detection on a SingleCells object

scdblfinder_sc.Rd

Cluster-aware doublet detection using engineered features and a gradient-boosted classifier. See Germain et al., F1000Research, 2022.

Usage

scdblfinder_sc(
  object,
  scdblfinder_params = params_scdblfinder(),
  return_features = FALSE,
  cells_to_use = NULL,
  group_by = NULL,
  streaming = NULL,
  seed = 42L,
  .verbose = TRUE
)

Arguments

object

SingleCells class.

scdblfinder_params

List. Parameters from params_scdblfinder().

return_features

Boolean. Shall the features used to train the classifier be returned.

cells_to_use

Optional string. Names of the cells to use for the run of the boosted doublet detection. Useful when you wish to run doublet detection on individual batches within your data. The object returned will be specifically using these cells.

group_by

Optional grouping variable. Useful if you want to run the method on a per-sample basis.

streaming

Optional Boolean. Shall the data be streamed in. Useful for larger data sets where you wish to avoid loading in the whole data. If NULL, will automatically detect.

seed

Integer. Seed for reproducibility.

.verbose

Boolean or integer. Controls verbosity and returns run times. FALSE -> quiet, TRUE or 1L -> normal verbosity, 2L -> detailed verbosity.

Value

An S3 object of class ScDblFinderRes containing:

predicted_doublets

Logical vector of doublet calls.

doublet_score

Numeric vector of classifier probabilities.

cxds_scores

Numeric vector of the cxds scores.

weighted

Numeric vector of the weighted scores.

threshold

The threshold used for calling.

cluster_labels

Integer vector of final cluster assignments.

detected_doublet_rate

Fraction of cells called as doublets.

with cell_indices stored as an attribute.

Examples

# cluster aware doublet calls from the gradient boosted classifier
sc <- demo_single_cells(prepped = FALSE)
scdblfinder_sc(
  sc,
  scdblfinder_params = params_scdblfinder(
    pca = list(no_pcs = 10L),
    n_genes = 25L,
    cxds_genes = 25L
  ),
  .verbose = FALSE
)
#> ScDblFinderRes: 500 cells, 14 doublets (2.8%)
#>   Threshold:        0.4883
#>   Score range:      [0.0361, 0.9485]
#>   Final clusters:   3
#>   Features available: FALSE

unlink(sc@dir_data, recursive = TRUE, force = TRUE)