graphClustering-methods: Graph-based clustering.

In edroaldo/fusca: Framework for Unified Single-Cell Analysis

Graph-based clustering.

Description

Identify clusters using graph-based model. Use jaccard or invlog similarity. Update graph, rawdata, ndata and sampTab.

Usage

graphClustering(
  object,
  assay.type = "RNA",
  k = 5,
  num.pcs,
  nn.type = "nng",
  sim.type = "jaccard",
  filename = "graph_subpopulations.gml"
)

## S4 method for signature 'CellRouter'
graphClustering(
  object,
  assay.type = "RNA",
  k = 5,
  num.pcs,
  nn.type = "nng",
  sim.type = "jaccard",
  filename = "graph_subpopulations.gml"
)

Arguments

object	CellRouter object
assay.type	character; the type of data to use.
k	numeric; number of nearest neighbors to build a k-nearest neighbors graph.
num.pcs	numeric; number of principal components that will define the space from where the kNN graph is identified. For example, if num.pcs = 10, the kNN graph will be created from a 10-dimensional PCA space.
nn.type	character; method to find the k-nearest neighbor graph. If 'nng', the code will use the nng function of the cccd package. If 'knn' or 'snn', it will use the functions from the bluster package, which are recommended for big data due to their efficiency.
sim.type	character; updates the kNN graph to encode cell-cell similarities using the jaccard or invlog methods.
filename	character; save .gml file containing the kNN graph.

Value

CellRouter object with the graph, rawdata, ndata, and sampTab slots updated.

edroaldo/fusca documentation built on March 1, 2023, 1:43 p.m.