Skip to contents

Doublet detection with Scrublet

scrublet_sc.Rd

This function implements the doublet detection from Scrublet, see Wolock, et al. Briefly, arteficial doublets are being generated from the data via random combination of initial cells. Highly variable genes (HVG) are being identified and the observed cells are being projected on a PCA space. Subsequently, the simulated doublets are being projected on the same PCA space given the same HVGs; kNN graphs are being generated and a kNN classifier is used to assign a probability that a given cell in the original data is a doublet. For more details, please check the publication.

Usage

scrublet_sc(
  object,
  scrublet_params = params_scrublet(),
  seed = 42L,
  streaming = NULL,
  cells_to_use = NULL,
  group_by = NULL,
  return_combined_pca = FALSE,
  return_pairs = FALSE,
  .verbose = TRUE
)

Arguments

object

SingleCells class.

scrublet_params

A list with the final scrublet parameters, see params_scrublet() for full details.

seed

Integer. Random seed.

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.

cells_to_use

Optional string. Names of the cells to use for the generation of the Scrublet. 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.

return_combined_pca

Boolean. Shall the PCA of the observed cells and simulated doublets be returned.

return_pairs

Boolean. Shall the pairs be returned.

.verbose

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

Value

A scrublet_res class that has with the following items:

  • predicted_doublets - Boolean vector indicating which observed cells predicted as doublets (TRUE = doublet, FALSE = singlet).

  • doublet_scores_obs - Numerical vector with the likelihood of being a doublet for the observed cells.

  • doublet_scores_sim - Numerical vector with the likelihood of being a doublet for the simulated cells.

  • doublet_errors_obs - Numerical vector with the standard errors of the scores for the observed cells.

  • z_scores - Z-scores for the observed cells. Represents: score - threshold / error.

  • threshold - Used threshold.

  • detected_doublet_rate - Fraction of cells that are called as doublet.

  • detectable_doublet_fraction - Fraction of simulated doublets with scores above the threshold.

  • overall_doublet_rate - Estimated overall doublet rate. Should roughly match the expected doublet rate.

  • pca - Optional PCA embeddings across the original cells and simulated doublets.

  • pair_1 - Optional index of the parent cell 1 of the simulated doublets.

  • pair_2 - Optional index of the parent cell 2 of the simulated doublets.

References

Wollock, et al., Cell Syst, 2020

Examples

# simulated doublet scoring on 500 synthetic cells
sc <- demo_single_cells(prepped = FALSE)
scrublet_sc(
  sc,
  scrublet_params = params_scrublet(
    pca = list(no_pcs = 10L),
    hvg = list(min_gene_var_pctl = 0.0),
    n_bins = 20L
  ),
  .verbose = FALSE
)
#> ScrubletRes: 500 cells, 36 doublets (7.2%)
#>   Threshold:              0.1215
#>   Detected doublet rate:  7.2%
#>   Detectable fraction:    97.7%
#>   Overall doublet rate:   7.4%
#>   Simulated doublets:     750

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