callClusters: Cluster cells using Louvain, Leiden or other graph-based...

In plantformatics/Socrates: Analysis of scATAC-seq data

Cluster cells using Louvain, Leiden or other graph-based methods implemented by Seurat

Description

Cluster cells using Louvain, Leiden or other graph-based methods implemented by Seurat

Usage

callClusters(
  obj,
  res = 0.4,
  k.near = 30,
  clustOB = "svd",
  cname = "LouvainClusters",
  min.reads = 50000,
  m.clst = 25,
  threshold = 5,
  umap1 = "umap1",
  umap2 = "umap2",
  e.thresh = 3,
  cleanCluster = T,
  cl.method = 1,
  svd_slotName = "PCA",
  umap_slotName = "UMAP",
  cluster_slotName = "Clusters",
  verbose = FALSE,
  ...
)

Arguments

obj	list, object containing 'PCA', 'UMAP', 'counts' and 'meta'. Required.
res	numeric, resolution for Louvain/Leiden clustering. Typical values range from 0 to 2. Defaults to 0.8.
k.near	numeric, number of nearest neighbors for graph building. Defaults to 20.
clustOB	character, which embedding to use for clustering. Can be one of c("umap", "svd"). It is highly recommended to use 'svd' for graph-based clustering. Defaults to "svd".
cname	character, column name in meta cluster IDs. Defaults to "LouvainClusters".
min.reads	numeric, minimum number of aggregated nSites (number of accessible peaks per cell) to consider a cluster as valid. Note: that this does not consider the total number of reads in each cell. Rather, we take the sum of column sums (colSum) for cells within a single cluster. Defaults to 5e4.
m.clst	numeric, minimum number of cells to consider a cluster as valid. Defaults to 25.
threshold	numeric, Z-score threshold to filter out cells greater than X distance to other cells on average. Uses a knn to find the top 50 nearest cells. See 'filterSingle' for more details.
umap1	character, column name of UMAP first component. Only important for sub-clustering.
umap2	character, column name of UMAP second component. Only important for sub-clustering.
e.thresh	numeric, Z-score threshold to filter out cells failing to co-localize with other cells apart of the same cluster. Set to 1e6 to skip cluster refining. This step may also be skipped by setting cleanCluster to FALSE. Default is 2.
cleanCluster	logical, whether or not to implement cluster refinement by removing cells that do not co-localize the the majority of cells apart of the same cluster. Uses the UMAP embedding for distance estimates.
cl.method	numeric, graph clustering algorithm. Numeric values range from 1-4, see Seurat::FindClusters for more details.
svd_slotName	character, character string for the SVD slot to use for clustering. Defaults to "PCA".
umap_slotName	character, character string for the UMAP slot to use for analysis. Defaults to "UMAP.
cluster_slotName	character, character string for naming the results of callClusters. Defaults to "Clusters".
verbose	logical. Defaults to FALSE.
...	additional arguments sent to Seurat::FindClusters
key	character, name of UMAP key when creating Seurat object. Defaults to "UMAP_".

plantformatics/Socrates documentation built on April 3, 2025, 1:02 p.m.