Data clustering

Giotto provides several options for clustering using the GiottoClass::clusterData() generic. Clustering for matrix-like data hooks into the bluster framework. Network community clustering type methods instead use Giotto-specific implementations that work on top of Giotto's calculated nearest neighbor networks.

# S4 method for class 'allMatrix,BlusterParam'
clusterData(
  x,
  param,
  cell_subset = NULL,
  feat_subset = NULL,
  cells_are_rows = TRUE,
  set_seed = TRUE,
  seed_number = 1234,
  output = c("data.table", "factor", "full"),
  ...
)

# S4 method for class 'allMatrix,NNClusParam'
clusterData(x, param, ...)

# S4 method for class 'exprObj,BlusterParam'
clusterData(
  x,
  param,
  cell_subset = NULL,
  feat_subset = NULL,
  set_seed = TRUE,
  seed_number = 1234,
  output = c("data.table", "factor", "full"),
  ...
)

# S4 method for class 'dimObj,BlusterParam'
clusterData(
  x,
  param,
  cell_subset = NULL,
  feat_subset = 1:10,
  set_seed = TRUE,
  seed_number = 1234,
  output = c("data.table", "factor", "full"),
  ...
)

# S4 method for class 'spatEnrObj,BlusterParam'
clusterData(
  x,
  param,
  cell_subset = NULL,
  feat_subset = NULL,
  set_seed = TRUE,
  seed_number = 1234,
  output = c("data.table", "factor", "full"),
  ...
)

# S4 method for class 'nnNetObj,NNClusParam'
clusterData(
  x,
  param,
  set_seed = TRUE,
  seed_number = 1234,
  output = c("data.table", "factor"),
  ...
)

# S4 method for class 'igraph,LouvainCommunityClusParam'
clusterData(
  x,
  param,
  set_seed = TRUE,
  seed_number = 1234,
  output = c("data.table", "factor")
)

# S4 method for class 'igraph,LouvainMultinetClusParam'
clusterData(
  x,
  param,
  set_seed = TRUE,
  seed_number = 1234,
  output = c("data.table", "factor")
)

# S4 method for class 'igraph,LeidenPythonClusParam'
clusterData(
  x,
  param,
  set_seed = TRUE,
  seed_number = 1234,
  output = c("data.table", "factor")
)

# S4 method for class 'igraph,LeidenIgraphClusParam'
clusterData(
  x,
  param,
  set_seed = TRUE,
  seed_number = 1234,
  output = c("data.table", "factor"),
  ...
)

# S4 method for class 'giotto,BlusterParam'
clusterData(
  x,
  param,
  what = c("expression", "dimension_reduction", "spatial_enrichment"),
  name = paste(class(param), "clusters", sep = "_"),
  dim_reduction_to_use = "pca",
  dim_reduction_name = "pca",
  expression_values = "normalized",
  spat_enr_name = NULL,
  spat_unit = NULL,
  feat_type = NULL,
  output = c("gobject", "data.table", "factor", "full"),
  .n = 1L,
  ...
)

# S4 method for class 'giotto,NNClusParam'
clusterData(
  x,
  param,
  name = paste(class(param), "clusters", sep = "_"),
  nn_network_to_use = "sNN",
  network_name = "sNN.pca",
  spat_unit = NULL,
  feat_type = NULL,
  output = c("gobject", "data.table", "factor"),
  .n = 1L,
  ...
)

Arguments

x

data object to cluster

param

S4 param class inheriting from bluster::BlusterParam or NNClusParam

cell_subset

numeric, logical, factor, or character vector of cells to use. (matrix-like clustering only)

feat_subset

numeric, logical, factor, or character vector of features or dimensions to use. (matrix-like clustering only)

cells_are_rows

logical. Transposes the input matrix when FALSE. Mainly here to highlight that clustering is performed on row-wise cells.

set_seed

logical, default = TRUE. Whether to set a seed for reproducibility.

seed_number

numeric, default = 1234. If set_seed = TRUE, seed number to use.

output

character. Type of output.

"factor" returns a numeric factor of clustering results, named by cell_ID.
"data.table" returns a two column (or one if no cell_IDs are present) with columns "cell_ID" and "cluster".
"full" returns the full output. It is equivalent to setting full = TRUE for bluster::clusterRows().
"gobject" can only be selected when x is a giotto object. The clustering results are appended to the cell metadata column designated by name param and the giotto object is returned.

...

additional params to pass to underlying algorithms, if any.

what

character. One of "expression", "dimension_reduction", or "spatial_enrichment" indicating what data in giotto object to cluster. (matrix-like clustering only)

name

character. When output = "gobject", the name to assign the column of clustering results in cell metadata.

dim_reduction_to_use

character. Type of dimension reduction to use.

dim_reduction_name

character. Name of dimension reduction results to use.

expression_values

character. Name of expression values to use

spat_enr_name

character. Name of spatial enrichment values to use.

spat_unit

character. Spatial unit to use. NULL will autodetect based on active spatial unit.

feat_type

character. Feature type to use. NULL will autodetect based on active feature type.

.n

integer. Used in recording object history (see update_giotto_params()). Number of stack frames back to record call of interest. Set to a large negative number (e.g. -1000) to avoid recording altogether (useful when there is another record call elsewhere in the stack you want to use.)

nn_network_to_use

character. Type of nearest neighbor network to use. Currently one of "sNN" or "kNN".

network_name

character. Name of nearest neighbor network to use.

Examples

x <- clusterParam("kmeans", centers = 2)
x@centers
#> [1] 2

m <- matrix(runif(9), nrow = 3)
clusterData(m, x)
#>    cluster
#>     <fctr>
#> 1:       1
#> 2:       1
#> 3:       2
# add ids
rownames(m) <- paste("id", seq_len(3), sep = "_")
clusterData(m, x)
#>    cell_ID cluster
#>     <char>  <fctr>
#> 1:    id_1       1
#> 2:    id_2       1
#> 3:    id_3       2

Arguments

See also

Examples