Calculate Leiden clustering for a Graph object in Seurat

Usage

FindClustersLeiden(
  object,
  graph.name = NULL,
  group.singletons = TRUE,
  resolution = c(0.8),
  min.cluster.size = 5,
  n_iterations = 2,
  verbose = TRUE,
  pythondir = "~/miniconda3/bin/python",
  partition_type = "RBConfigurationVertexPartition"
)

Arguments

object: Seurat object
graph.name: Name of Graph slot in object to use for Leiden clustering
group.singletons: Group singletons into nearest cluster. If FALSE, the clusters will remain as single-member clusters.
resolution: Vector of resolution to input into Leiden
min.cluster.size: Minimum cluster size trimmed after clustering (not currently used)
verbose: Control verbosity
pythondir: Specified director for python binary that has the leidenalg library installed
partition_type: The partition_type to specify for Leiden.
n.iterations: Number of iterations to run Leiden clustering for

Value

Returns a Seurat object with the leiden clusterings stored as object@meta.data columns

Details

code is adapted from @immunogenomics/singlecellmethods with some tweaks and flexibility added to match the arguments allowed by the leidenalg package. In essence, we are using reticulate to run leidenalg.find_partition(graph.name, partition_type) as specified here https://leidenalg.readthedocs.io/en/stable/intro.html. Arguments are input into the find_partition function here: https://leidenalg.readthedocs.io/en/stable/reference.html#leidenalg.find_partition. Singletons will be excluded for now, but ideally can be grouped in a step afterwards...

References

code is adapted from @immunogenomics/singlecellmethods