Register a clustering method for use by name

Adds a clustering backend to the package registry so it can be selected with clusterGeneSets(algorithm = "<name>"). Registering is optional – a gsClusterMethod (or a bare function) can also be passed straight to clusterGeneSets().

Usage

registerClusterMethod(
  name,
  fn,
  description = NA_character_,
  packages = character(0),
  usesResolution = TRUE,
  overwrite = FALSE
)

Arguments

name: Name to register the method under.
fn: Either a gsClusterMethod (from clusterMethod()) or a function following the same contract; a bare function is wrapped automatically.
description: One-line description; ignored when fn is already a gsClusterMethod.
packages: Packages the method requires; ignored when fn is already a gsClusterMethod.
usesResolution: Whether the method's output depends on resolution; ignored when fn is already a gsClusterMethod. See clusterMethod().
overwrite: Whether to replace an already-registered method of the same name. Defaults to FALSE.

Value

The registered gsClusterMethod, invisibly.

Examples

registerClusterMethod(
  "leiden_cpm",
  function(snn, resolution, verbose = FALSE, ...) {
    g <- snnToIgraph(snn)
    igraph::membership(igraph::cluster_leiden(
      g,
      objective_function = "CPM",
      resolution = resolution,
      weights = igraph::E(g)$weight
    ))
  },
  description = "igraph Leiden with the CPM objective",
  packages = "igraph",
  overwrite = TRUE
)
listClusterMethods()
#>                  name
#> 1           apcluster
#> 2          binary_cut
#> 3       igraph_leiden
#> 4      igraph_louvain
#> 5              kmeans
#> 6              leiden
#> 7          leiden_cpm
#> 8             louvain
#> 9  louvain.multilevel
#> 10                slm
#>                                                                                               description
#> 1                                Affinity propagation (apcluster::apcluster) on the SNN similarity matrix
#> 2                                             simplifyEnrichment::binary_cut on the SNN similarity matrix
#> 3                                          igraph::cluster_leiden on the SNN graph (modularity objective)
#> 4                                                                igraph::cluster_louvain on the SNN graph
#> 5  k-means on SNN similarity profiles; k chosen by elbow/knee on the WSS curve (as in simplifyEnrichment)
#> 6                                                Seurat::FindClusters, Leiden via leidenalg (algorithm 4)
#> 7                                                                    igraph Leiden with the CPM objective
#> 8                                                    Seurat::FindClusters, original Louvain (algorithm 1)
#> 9                                  Seurat::FindClusters, Louvain with multilevel refinement (algorithm 2)
#> 10                                                 Seurat::FindClusters, smart local moving (algorithm 3)
#>              packages uses_resolution
#> 1           apcluster           FALSE
#> 2  simplifyEnrichment           FALSE
#> 3              igraph            TRUE
#> 4              igraph            TRUE
#> 5                               FALSE
#> 6      Seurat, leiden            TRUE
#> 7              igraph            TRUE
#> 8              Seurat            TRUE
#> 9              Seurat            TRUE
#> 10             Seurat            TRUE

geneSetsList <- initializeList(exampleGeneSets, exampleEnrichment,
                               verbose = FALSE)
gsListObject <- clusterGeneSets(geneSetsList, algorithm = "leiden_cpm",
                                resolution = 0.05, verbose = FALSE)

Usage

Arguments

Value

See also

Examples