downsample matrix by cluster or completely random

Usage

downsample_matrix(
  mat,
  n = 1,
  keep_cluster_proportions = TRUE,
  metadata = NULL,
  cluster_col = "cluster"
)

Arguments

mat: expression matrix
n: number per cluster or fraction to keep
keep_cluster_proportions: whether to subsample
metadata: data.frame or vector containing cluster assignments per cell. Order must match column order in supplied matrix. If a data.frame provide the cluster_col parameters.
cluster_col: column in metadata with cluster number

Value

new smaller mat with less cell_id columns

Examples

set.seed(42)
mat <- downsample_matrix(
    mat = pbmc_matrix_small,
    metadata = pbmc_meta$classified,
    n = 10,
    keep_cluster_proportions = TRUE
)
mat[1:3, 1:3]
#> 3 x 3 sparse Matrix of class "dgCMatrix"
#>        ACGTCGCTTCTCAT TTACCATGTTGTGG CCTGACTGTGTCTT
#> PPBP                .       .              .       
#> LYZ                 .       2.633919       2.592335
#> S100A9              .       .              .