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.
Examples
utils::data("pbmc_small", package = "SeuratObject")
mat <- SeuratObject::LayerData(
pbmc_small,
assay = "RNA",
layer = "counts"
)
fs <- gini.rank.fs(
mat,
num_threads = 1
)