Skip to contents

Pipeline step: generate meta cells

step_metacells_sc.Rd

Wraps the meta cell generators as an ScStep. Unlike the other steps this one changes the class of the object: it takes a SingleCells or SingleCellsSubset and returns a MetaCells(). Steps that follow it need MetaCells methods, which validate_pipeline() checks up front.

Combined with apply_pipeline_per_group() this gives you per-group pre-processing (HVG, PCA, batch correction within a patient, kNN) followed by source-pure meta cells, which you then hand to merge_meta_cells().

Usage

step_metacells_sc(method = c("bootstrapped", "seacells", "supercells"), ...)

Arguments

method

String. One of c("bootstrapped", "seacells", "supercells").

...

Arguments passed on to the generator, e.g. sc_meta_cell_params, target_size or .verbose.

Value

An ScStep.

Examples

# per group pre-processing that ends on source-pure meta cells
sc <- demo_single_cells(prepped = FALSE)
pipeline <- step_hvg_sc(hvg_no = 30L, .verbose = FALSE) %>>%
  step_pca_sc(no_pcs = 10L, .verbose = FALSE) %>>%
  step_neighbours_sc(.verbose = FALSE) %>>%
  step_metacells_sc(
    "bootstrapped",
    sc_meta_cell_params = params_sc_bt_metacells(target_no_metacells = 10L),
    .verbose = FALSE
  )

per_group <- apply_pipeline_per_group(pipeline, sc, group_col = "cell_grp")
merge_meta_cells(per_group, .verbose = FALSE)
#> Single cell experiment (Meta Cells).
#>   Meta cell method: meta_cells_hdwgcna
#>   Merged: TRUE
#>   No meta cells: 30
#>   No genes: 50
#>   No cells aggregated: 297
#>   No obs rows in source: 500
#>   HVG calculated: FALSE
#>   PCA calculated: FALSE
#>   Other embeddings: none
#>   KNN generated: FALSE
#>   SNN generated: FALSE
#>   Stale artefacts: none

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