Skip to contents

Identifies specific and non-specific features based on the Gini inequality coefficient computed on ranked expression values.

Usage

gini.rank.fs(mat, gini_thresh = 0.5, noise_thresh = NULL, num_threads = -1)

Arguments

mat

A matrix with features as rows and cells as columns.

gini_thresh

Numeric; Gini threshold for selecting specific features.

noise_thresh

Integer; minimum number of cells required for a feature to be retained.

num_threads

Integer; number of threads to use. -1 uses all available cores.

Value

A list containing specific_features, non_specific_features, and no_occurrence.

Details

This method removes globally low-ranked features while retaining features with consistently high ranks across cells for downstream analysis.

See also

Examples

utils::data("pbmc_small", package = "SeuratObject")

mat <- SeuratObject::LayerData(
  pbmc_small,
  assay = "RNA",
  layer = "counts"
)

fs <- gini.rank.fs(
  mat,
  num_threads = 1
)