Skip to contents

Wrapper function for parameters for HotSpot

params_sc_hotspot.Rd

weighted_graph controls how the kNN distances become edge weights. The default of FALSE follows the reference implementation: the distances only decide who is a neighbour and every retained edge weighs one. Set it to TRUE for the Gaussian kernel, whose width is the ceil(k / neighborhood_factor)-th neighbour distance.

Usage

params_sc_hotspot(
  model = c("danb", "normal", "bernoulli"),
  normalise = TRUE,
  weighted_graph = FALSE,
  neighborhood_factor = 3,
  knn = list()
)

Arguments

model

String. Model to use for modelling the GEX. One of c("danb", "normal", "bernoulli"). Defaults to "danb".

normalise

Boolean. Shall the data be normalised. Defaults to TRUE.

weighted_graph

Boolean. Shall the Gaussian kernel be applied to the neighbour distances. Defaults to FALSE.

neighborhood_factor

Numeric. Kernel width is the ceil(k / neighborhood_factor)-th neighbour distance. Only read when weighted_graph = TRUE. Defaults to 3.0.

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:

  • model - String. Model to use for modelling the GEX. One of c("danb", "normal", "bernoulli"). Defaults to "danb".

  • normalise - Boolean. Shall the data be normalised. Defaults to TRUE.

  • weighted_graph - Boolean. Shall the Gaussian kernel be applied to the neighbour distances. Defaults to FALSE.

  • neighborhood_factor - Numeric. Kernel width is the ceil(k / neighborhood_factor)-th neighbour distance. Only read when weighted_graph = TRUE. Defaults to 3.0.

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

References

DeTomaso and Yosef, Cell Systems, 2021