clusterSignalMatrices: clusters the regions of a (set of) signal matrices.

Usage

clusterSignalMatrices(
  ml,
  k,
  scaleRows = FALSE,
  scaleCols = FALSE,
  use = c("enrich", "full", "max", "center"),
  by = rep(1L, length(ml)),
  assay = 1L,
  trim = c(0.05, 0.95),
  nstart = 3,
  ...
)

Arguments

ml: A named list of signal matrices or an EnrichmentSE object as produced by signal2Matrix
k: The number of clusters to generate
scaleRows: Logical; whether to scale rows for clustering
scaleCols: Logical; whether to scale columns (i.e. signals/samples)
use: What values to use for clustering. By default, uses enriched_score. Other options are 'full' (uses the full signal for clustering), 'max' (uses the maximum value in the region), or 'center' (use the value at the center of the region).
by: Optional factor/character/integer vector of the same length as `ml`. When scaling rows, this can be used to indicate which rows should be scaled together.
assay: Assay to use (ignored unless `ml` is an ESE object)
trim: Values to trim (applied individually for each signal matrix)
nstart: Number of starts for k-means clustering
...: Passed to `kmeans`

Value

If `k` is of length 1, a vector of cluster labels, corresponding to the rows of `ml`. If `length(k)>1`, a list of two data.frames containing: 1) the cluster labels at the different resolutions, and 2) the variance explained by clusters at each resolution.

Examples

data(exampleESE)
rowData(exampleESE)$cluster <- clusterSignalMatrices(exampleESE, 3)
#>   ~82% of the variance explained by clusters
# we could plot the data clustered:
plotEnrichedHeatmaps(exampleESE, row_split="cluster")

# we could also try with different values of k:
cl <- clusterSignalMatrices(exampleESE, 2:5)
cl$varExplained
#>   k varExplained
#> 2 2           79
#> 3 3           87
#> 4 4           89
#> 5 5           90