Skip to contents

Run RCTD spatial deconvolution

Source: R/RunRCTD.R
RunRCTD.Rd

Estimate spot-level cell type proportions from a spatial Seurat object using a single-cell Seurat reference and spacexr RCTD.

Usage

RunRCTD(
  srt,
  reference,
  reference_label = "celltype",
  assay = NULL,
  reference_assay = NULL,
  layer = "counts",
  reference_layer = "counts",
  features = NULL,
  image = NULL,
  coord.cols = c("x", "y"),
  rctd_mode = c("full", "multi", "doublet"),
  max_cores = 1,
  min_cells = 25,
  prefix = "RCTD",
  store_results = TRUE,
  round_counts = TRUE,
  create_rctd_params = list(),
  run_rctd_params = list(),
  verbose = TRUE,
  ...,
  coordinate_space = c("raw", "legacy_display"),
  tool_name = "RCTD"
)

Arguments

srt

Spatial Seurat object used as the RCTD query.

reference

Reference Seurat object containing annotated single cells.

reference_label

Metadata column in reference with cell type labels.

assay

Assay used in srt. If NULL, the default assay is used.

reference_assay

Assay used in reference.

layer, reference_layer

Assay layers used for spatial and reference raw counts.

features

Features used for RCTD. If NULL, shared features are used.

image

Name of the Seurat spatial image used to recover coordinates when coord.cols are not available.

coord.cols

Metadata coordinate columns used when no image coordinate source is requested or available.

rctd_mode

RCTD mode passed to spacexr. "full" is the default for Visium spot deconvolution.

max_cores

Number of cores passed to spacexr.

min_cells

Minimum number of reference cells required for each cell type. Old spacexr RCTD requires at least 25 cells per type.

prefix

Prefix for metadata columns.

store_results

Whether to store detailed RCTD results in srt@tools.

round_counts

Whether to round non-integer counts to the nearest integer before passing data to spacexr. RCTD requires integer count matrices; this defaults to TRUE so bundled example data with scaled non-integer reference counts can run directly.

create_rctd_params

Additional parameters passed to spacexr::createRctd() or spacexr::create.RCTD().

run_rctd_params

Additional parameters passed to spacexr::runRctd() or spacexr::run.RCTD().

verbose

Whether to print the message. Default is TRUE.

...

Additional parameters passed to the RCTD run step.

coordinate_space

Coordinate space used for distance-sensitive input. The default is raw acquisition coordinates, so backend distances use raw coordinate units. Use "legacy_display" explicitly to reproduce the display-scaled coordinates used before scop 0.9.0.

tool_name

Name used to store the plain result bundle in srt@tools.

Value

A Seurat object with RCTD proportion columns in metadata and dominant cell type summaries. When store_results = TRUE, detailed results are also stored in srt@tools[[tool_name]].

Examples

data(visium_human_pancreas_sub)
spatial <- visium_human_pancreas_sub
rctd_weights <- data.frame(
  RCTD_prop_Ductal = seq(0.75, 0.15, length.out = ncol(spatial)),
  RCTD_prop_Endocrine = seq(0.15, 0.65, length.out = ncol(spatial)),
  RCTD_prop_Stromal = 0.10,
  row.names = colnames(spatial)
)
rctd_weights <- rctd_weights / rowSums(rctd_weights)
spatial <- Seurat::AddMetaData(spatial, rctd_weights)
spatial$RCTD_dominant_type <- sub(
  "^RCTD_prop_",
  "",
  colnames(rctd_weights)[max.col(rctd_weights)]
)
spatial$RCTD_max_prop <- apply(rctd_weights, 1, max)

SpatialSpotPlot(
  spatial,
  group.by = "RCTD_dominant_type",
  overlay_image = FALSE,
  coord.cols = c("x", "y")
)

SpatialSpotPlot(
  spatial,
  group.by = "RCTD_dominant_type",
  plot_type = "pie",
  overlay_image = FALSE,
  coord.cols = c("x", "y")
)


data(panc8_sub)
spatial <- spatial[, seq_len(120)]
#> Warning: Not validating Centroids objects
#> Warning: Not validating Centroids objects
#> Warning: Not validating FOV objects
#> Warning: Not validating FOV objects
#> Warning: Not validating FOV objects
#> Warning: Not validating FOV objects
#> Warning: Not validating FOV objects
#> Warning: Not validating FOV objects
#> Warning: Not validating Seurat objects
reference <- panc8_sub[, panc8_sub$celltype %in% c("ductal", "alpha", "beta")]
reference <- Seurat::FindVariableFeatures(reference, nfeatures = 300, verbose = FALSE)
features_use <- intersect(
  SeuratObject::VariableFeatures(reference),
  rownames(spatial)
)

spatial <- RunRCTD(
  srt = spatial,
  reference = reference,
  reference_label = "celltype",
  assay = "Spatial",
  reference_assay = "RNA",
  layer = "counts",
  reference_layer = "counts",
  features = features_use,
  rctd_mode = "full",
  max_cores = 1,
  min_cells = 25,
  prefix = "RCTD",
  verbose = FALSE
)
#> Warning: Reference: some nUMI values are less than min_UMI = 100, and these cells will be removed. Optionally, you may lower the min_UMI parameter.
#> Begin: process_cell_type_info
#> process_cell_type_info: number of cells in reference: 694
#> process_cell_type_info: number of genes in reference: 97
#> 
#> ductal   beta  alpha 
#>    211    304    179 
#> End: process_cell_type_info
#> create.RCTD: getting regression differentially expressed genes: 
#> get_de_genes: ductal found DE genes: 24
#> get_de_genes: beta found DE genes: 7
#> get_de_genes: alpha found DE genes: 51
#> get_de_genes: total DE genes: 82
#> create.RCTD: getting platform effect normalization differentially expressed genes: 
#> get_de_genes: ductal found DE genes: 27
#> get_de_genes: beta found DE genes: 8
#> get_de_genes: alpha found DE genes: 55
#> get_de_genes: total DE genes: 90
#> fitBulk: decomposing bulk
#> chooseSigma: using initial Q_mat with sigma =  1
#> Likelihood value: 565.708541011282
#> Sigma value:  0.84
#> Likelihood value: 554.037735497346
#> Sigma value:  0.69
#> Likelihood value: 544.81606233551
#> Sigma value:  0.61
#> Likelihood value: 540.997124129871
#> Sigma value:  0.53
#> Likelihood value: 538.284148657321
#> Sigma value:  0.45
#> Likelihood value: 537.061480361226
#> Sigma value:  0.44
#> Likelihood value: 537.038433383715
#> Sigma value:  0.44

rctd_cols <- grep("^RCTD_prop_", colnames(spatial@meta.data), value = TRUE)
SpatialSpotPlot(
  spatial,
  group.by = "RCTD_dominant_type",
  overlay_image = FALSE,
  coord.cols = c("x", "y"),
  theme_use = "theme_scop"
)

SpatialSpotPlot(
  spatial,
  group.by = rctd_cols[1:min(3, length(rctd_cols))],
  palette = "Spectral",
  overlay_image = FALSE,
  coord.cols = c("x", "y"),
  theme_use = "theme_scop"
)