Skip to contents

Wrapper function for parameters for SuperCell generation

params_sc_supercell.Rd

Wrapper function for parameters for SuperCell generation

Usage

params_sc_supercell(
  walk_length = 3L,
  graining_factor = 20,
  use_kernel = TRUE,
  k_ith = NULL,
  max_support = NULL,
  knn = list()
)

Arguments

walk_length

Integer. Walk length for the Walktrap algorithm. Defaults to 3L.

graining_factor

Numeric. Graining level of data (proportion of number of single cells in the initial dataset to the number of metacells in the final dataset). (One meta cell per 20 cells.) Defaults to 20.0.

use_kernel

Boolean. Shall a kernel function akin to MAGIC be applied akin to the approach in SuperCell2, see Hérault, et al., bioRxiv, 2026 and van Dijk, et al., Cell, 2018. Defaults to TRUE.

k_ith

Integer or NULL. The k-ith neighbour to use for the kernel. Defaults to NULL.

max_support

Integer or NULL. Caps each cell's walk-probability vector to its top entries by mass, bounding memory at ~max_support * n_cells on large data. Makes the result an approximation. NULL (default) keeps the walks exact. Defaults to NULL.

knn

List. Optional overrides for kNN parameters. See params_knn_defaults() for available parameters: k, knn_method, ann_dist, search_budget, n_trees, delta, diversify_prob, ef_budget, extract_knn, m, ef_construction, ef_search, n_list and n_probe. See params_knn_defaults() for the available elements. Defaults to list().

Value

A named list with the following elements:

  • walk_length - Integer. Walk length for the Walktrap algorithm. Defaults to 3L.

  • graining_factor - Numeric. Graining level of data (proportion of number of single cells in the initial dataset to the number of metacells in the final dataset). (One meta cell per 20 cells.) Defaults to 20.0.

  • use_kernel - Boolean. Shall a kernel function akin to MAGIC be applied akin to the approach in SuperCell2, see Hérault, et al., bioRxiv, 2026 and van Dijk, et al., Cell, 2018. Defaults to TRUE.

  • k_ith - Integer or NULL. The k-ith neighbour to use for the kernel. Defaults to NULL.

  • max_support - Integer or NULL. Caps each cell's walk-probability vector to its top entries by mass, bounding memory at ~max_support * n_cells on large data. Makes the result an approximation. NULL (default) keeps the walks exact. Defaults to NULL.

  • The elements of params_knn_defaults(), overridden by knn, spliced in at this position.