Take one cluster from an existing clustering and partition it further by
community detection on the sub-graph of its cells, similar to
Seurat::FindSubCluster. New sub-clusters are labelled <cluster>_1,
<cluster>_2, ... (largest first).
Usage
getSubClusters(
spe,
cluster,
resolution = 1,
nbrs_name = NULL,
method = c("leiden", "louvain"),
cluster_name = "cluster",
new_name = cluster_name,
relabel = FALSE,
min_size = NULL,
start_from = NULL,
sep = "_",
seed = 1,
...
)Arguments
- spe
A SpatialExperiment object.
- cluster
The cluster label to sub-cluster (e.g. 2 or "2").
- resolution
Resolution for the sub-clustering. Higher gives more sub-clusters. See cluster_leiden and cluster_louvain.
- nbrs_name
Name of neighbour list to use. If NULL, uses the newest one in spe@metadata$nbrs$cell. findNbrsSNN must have been run first.
- method
Clustering method. Options are leiden and louvain.
- cluster_name
Name of the cluster column in colData to sub-cluster.
- new_name
Name of the column to store the result. Defaults to overwriting cluster_name.
- relabel
Logical. If TRUE, renumber all clusters and sub-clusters by size (largest first) after sub-clustering, replacing the
<cluster>_<n>labels with plain numbers. "unassigned" is kept as the last level. Defaults to FALSE.- min_size
Sub-clusters with at most this many cells are merged into the larger sub-clusters (graph connection first, then nearest centroid), as in getClusters. Defaults to NULL, which uses either 5 or 0.01\ the cells, whichever is smaller.
- start_from
Integer at which numbering starts, for both the sub-cluster suffixes and the relabel renumbering. Defaults to NULL, which infers the base (0- or 1-based) from the existing labels.
- sep
Separator between a cluster and its sub-cluster number. Default "_".
- seed
Seed for clustering.
- ...
Other clustering arguments for cluster_leiden or cluster_louvain.
Details
The sub-graph of the target cluster's cells is induced from the SNN graph and clustered on its own. Cells that are weakly connected within the cluster may form small or singleton sub-clusters; lower the resolution if this is not wanted.
Examples
data("xenium_bc_spe")
spe <- normalizeAssay(spe)
spe <- runPCA(spe)
#> Genes with 0 variance are excluded: ENSG00000135218 NegControlProbe_00002 NegControlCodeword_0504 NegControlCodeword_0509 NegControlCodeword_0510 NegControlCodeword_0511 NegControlCodeword_0512 NegControlCodeword_0516 NegControlCodeword_0517 NegControlCodeword_0518 NegControlCodeword_0519 NegControlCodeword_0520 NegControlCodeword_0522 NegControlCodeword_0526 NegControlCodeword_0527 NegControlCodeword_0530 NegControlCodeword_0536 NegControlCodeword_0537 BLANK_0030 BLANK_0163 BLANK_0165 BLANK_0212 BLANK_0221 BLANK_0230 BLANK_0237 BLANK_0311 BLANK_0361 BLANK_0365 BLANK_0382 BLANK_0384 BLANK_0387 BLANK_0388 BLANK_0391 BLANK_0393 BLANK_0397 BLANK_0399 BLANK_0404 BLANK_0406 BLANK_0410 BLANK_0411 BLANK_0418 BLANK_0425 BLANK_0432 BLANK_0447
spe <- findNbrsSNN(spe, dimred = "PCA")
#> [1] "Getting K-nearest neighbour"
#> [1] "Getting shared nearest neighbour"
spe <- getClusters(spe, resolution = 0.5)
#> Reassigned all 1 cells from clusters with <= 1 cells (0 by graph, 1 by distance).
# Split cluster 2 into 2_1, 2_2, ...
spe <- getSubClusters(spe, cluster = 2, resolution = 0.5)
#> Split cluster '2' into 4 sub-clusters.