Detect rare cells with native or optional rare-cell backends

Usage

sn_detect_rare_cells(
  object,
  method = c("gini", "sccad", "sca", "gapclust", "challenging_groups"),
  group = NULL,
  reduction = .sn_default_metric_reduction(object),
  dims = NULL,
  assay = "RNA",
  layer = "data",
  nfeatures = 200,
  min_cells = 3,
  max_fraction = 0.1,
  threshold = NULL,
  k = 20,
  seed = 717,
  sccad_python = NULL,
  sccad_script = NULL,
  sccad_normalization = FALSE,
  sccad_rare_h = 0.01,
  sccad_merge_h = 0.3,
  sccad_overlap_h = 0.7,
  gapclust_k = 200,
  sca_python = NULL,
  sca_n_comps = 20,
  sca_iters = 3,
  sca_nbhd_size = 15,
  sca_model = "wilcoxon"
)

Arguments

object: A Seurat object.
method: Rare-cell method. Supported values are "gini", "sccad", "sca", "gapclust", and "challenging_groups".
group: Optional metadata column used with method = "challenging_groups".
reduction: Reduction used by graph-based methods. Defaults to "harmony" when present, otherwise "pca".
dims: Optional embedding dimensions to use.
assay: Assay used to extract expression values.
layer: Layer used to extract expression values for score-based methods.
nfeatures: Number of rare-aware genes to use for score construction.
min_cells: Minimum number of cells a gene must be detected in before it is considered by score-based methods.
max_fraction: Maximum expressing-cell fraction for score-based rare genes.
threshold: Optional explicit threshold on the rare-cell score. When NULL, the function uses the upper-IQR rule.
k: Number of neighbors for graph-based methods.
seed: Random seed used by stochastic backends.
sccad_python: Optional Python executable used by the scCAD backend.
sccad_script: Optional path to the upstream scCAD.py script.
sccad_normalization: Whether scCAD should normalize the provided matrix internally.
sccad_rare_h: Rare threshold passed to scCAD.
sccad_merge_h: Merge threshold passed to scCAD.
sccad_overlap_h: Overlap threshold passed to scCAD.
gapclust_k: Upper limit of the minor-cluster size used by GapClust.
sca_python: Optional Python executable used by the SCA backend.
sca_n_comps: Number of SCA components used before rarity scoring.
sca_iters: Number of SCA iterations.
sca_nbhd_size: Neighborhood size passed to SCA.
sca_model: Scoring model passed to SCA.

Value

A data frame with one row per cell, including a rare_score column and a logical rare_cell flag.

Examples

if (FALSE) { # \dontrun{
data("pbmc_small", package = "Shennong")
rare_tbl <- sn_detect_rare_cells(pbmc_small, method = "gini")
head(rare_tbl)
} # }