Compute clusters for multiple resolutions and saves in the metadata the clustering result that reaches the maximum NMI and/or ARI value for a given cell-type label variable.
Usage
FindOptimalClusters(
object,
graph.name = NULL,
cell.var = NULL,
cluster.name = "{graph.name}_{cell.var}_{metric}",
modularity.fxn = 1,
initial.membership = NULL,
node.sizes = NULL,
resolutions = seq(from = 0.1, to = 2, by = 0.1),
optimisation.metric = c("nmi", "ari"),
method = "matrix",
algorithm = 1,
n.start = 10,
n.iter = 10,
random.seed = 0,
group.singletons = TRUE,
temp.file.location = NULL,
edge.file.name = NULL,
verbose = TRUE,
...
)Arguments
- object
a Seurat object
- graph.name
the name of the knn graph to score.
- cell.var
The name(s) of the column(s) with cell type label variable (must be in the object metadata). Multiple column names are accepted
- cluster.name
a (optionally 'glue') string used as the new metadata column name (see Details section)
- modularity.fxn
Modularity function (1 = standard; 2 = alternative).
- initial.membership, node.sizes
Parameters to pass to the Python leidenalg function.
- resolutions
the resolutions to compute clusters for
- optimisation.metric
one of "nmi" or "ari" or both (default). The metric(s) to use to check clustering results against the
cell.var.- method
Method for running leiden (defaults to matrix which is fast for small datasets). Enable method = "igraph" to avoid casting large data to a dense matrix.
- algorithm
Algorithm for modularity optimization (1 = original Louvain algorithm; 2 = Louvain algorithm with multilevel refinement; 3 = SLM algorithm; 4 = Leiden algorithm). Leiden requires the leidenalg python.
- n.start
Number of random starts.
- n.iter
Maximal number of iterations per random start.
- random.seed
Seed of the random number generator.
- group.singletons
Group singletons into nearest cluster. If FALSE, assign all singletons to a "singleton" group
- temp.file.location
Directory where intermediate files will be written. Specify the ABSOLUTE path.
- edge.file.name
Edge file to use as input for modularity optimizer jar.
- verbose
Print output
- ...
Arguments passed to other methods
Details
cluster.name can use the 'glue' syntax to avoid overwriting metadata
columns when multiple metrics and/or cell-type label variables are provided.
But it can be ignored otherwise. Injectable variables are "graph.name",
"cell.var" and "metric" (not "optimisation.metric"). They
must be flanked by single curly brackets ("{" and "}").
For instance, if you prefer to name the clusters with an integration instead
of the graph.name, don't use the glue syntax for it (e.g.
"combat_{cell.var}_{metric}" would work, but any of
"{combat}_{cell.var}_{metric}" or
"{integration}_{cell.var}_{metric}" would throw an error)
References
Luecken, M. D., Büttner, M., Chaichoompu, K., Danese, A., Interlandi, M., Mueller, M. F., Strobl, D. C., Zappia, L., Dugas, M., Colomé-Tatché, M. & Theis, F. J. Benchmarking atlas-level data integration in single-cell genomics. Nat Methods 19, 41–50 (2021). DOI
See also
FindClusters, ScoreARI and
ScoreNMI