Skip to contents

Run BANKSY spatial clustering

Source: R/RunBANKSY.R
RunBANKSY.Rd

Build neighborhood-augmented BANKSY features from a spatial Seurat object and store spatial domain or microenvironment clusters in metadata.

Usage

RunBANKSY(
  object,
  assay = NULL,
  layer = "data",
  features = NULL,
  image = NULL,
  coord.cols = c("col", "row"),
  lambda = 0.2,
  k_geom = 15,
  M = 1,
  npcs = 20,
  use_agf = FALSE,
  algo = "leiden",
  k_neighbors = 50,
  resolution = 0.6,
  group = NULL,
  seed = 1,
  compute_banksy_params = list(),
  run_pca_params = list(),
  cluster_banksy_params = list(),
  cluster_source = NULL,
  cluster_colname = "BANKSY_cluster",
  tool_name = "BANKSY",
  store_results = TRUE,
  verbose = TRUE,
  coordinate_space = c("raw", "legacy_display"),
  srt = NULL,
  sample.by = NULL
)

Arguments

object

A Seurat object.

assay

Expression assay used by BANKSY. When NULL, the assay linked to the selected image is used first, followed by Spatial, RNA, and the object's default assay.

layer

Assay layer used as BANKSY input.

features

Optional features to use. If NULL, all assay features are used after zero-count filtering.

image

Spatial image used to recover coordinates. A single available image is selected automatically; choose one explicitly when several exist, or provide a named sample-to-image map with sample.by.

coord.cols

Metadata coordinate columns used when no image coordinate source is available. The default is resolved from common coordinate names.

lambda

BANKSY spatial weighting parameter.

k_geom

Unitless number of spatial neighbors used by BANKSY.

M

Highest azimuthal Fourier harmonic passed to BANKSY.

npcs

Number of principal components to compute.

use_agf

Whether to use azimuthal Gabor filters.

algo

Clustering algorithm passed to Banksy::clusterBanksy().

k_neighbors

Number of neighbors for graph clustering.

resolution

Graph clustering resolution.

group

Optional metadata column passed to BANKSY for multi-sample scaling. It is an algorithm parameter and does not split sample fits.

seed

Optional seed for PCA and clustering.

compute_banksy_params

Additional parameters passed to Banksy::computeBanksy().

run_pca_params

Additional parameters passed to Banksy::runBanksyPCA().

cluster_banksy_params

Additional parameters passed to Banksy::clusterBanksy().

cluster_source

Optional BANKSY colData column to copy. If NULL, the first cluster name reported by Banksy::clusterNames() is used when available.

cluster_colname

Metadata column used for BANKSY clusters.

tool_name

Name used to store detailed results in srt@tools.

store_results

Whether to store detailed BANKSY results in object@tools; cluster assignments are still written to metadata when FALSE; a lightweight index retains the result column and analyzed cells.

verbose

Whether to print the message. Default is TRUE.

coordinate_space

Coordinate space used for BANKSY spatial input. The default is raw acquisition coordinates, so geometry and distance weighting use raw coordinate units. Use "legacy_display" explicitly to reproduce the display-scaled coordinates used before scop 0.9.0.

srt

Deprecated alias for object; supply exactly one of the two. It will be removed in scop 1.0.0.

sample.by

Optional metadata column for independent per-sample fits. Combined labels are prefixed with the sample name; domains are not aligned across samples.

Value

A Seurat object with BANKSY clusters in metadata. When store_results = TRUE, detailed results are stored in srt@tools[[tool_name]]; when FALSE, clusters remain in metadata and only their lightweight retrieval index is retained in the tool entry. Both modes also retain one private metadata identity column per result, allowing GetSpatialResult() to follow subsets and renamed cells. Preserve this column; its name is recorded in result_index$cell_id_colname.

See also

GetSpatialResult()

Examples

data(visium_human_pancreas_sub)
keep_spots <- unique(round(seq(1, ncol(visium_human_pancreas_sub), length.out = 400)))
spatial <- visium_human_pancreas_sub[, keep_spots]
if (FALSE) { # \dontrun{
spatial <- RunBANKSY(
  spatial,
  layer = "counts",
  features = rownames(spatial)[1:200],
  resolution = 0.6,
  verbose = FALSE
)
SpatialSpotPlot(
  spatial,
  group.by = "BANKSY_cluster",
  pt.size = 1.5
)
} # }