Load CellGraphs — LoadCellGraphs • pixelatorR

Loads a list of CellGraph objects.

Usage

LoadCellGraphs(object, ...)

# S3 method for class 'FileSystemDataset'
LoadCellGraphs(
  object,
  cells,
  load_as = c("bipartite", "Anode", "linegraph"),
  add_marker_counts = TRUE,
  data_type = c("PNA", "MPX"),
  chunk_size = 10,
  verbose = TRUE,
  ...
)

# S3 method for class 'MPXAssay'
LoadCellGraphs(
  object,
  cells = colnames(object),
  load_as = c("bipartite", "Anode", "linegraph"),
  add_marker_counts = TRUE,
  add_layouts = FALSE,
  force = FALSE,
  chunk_size = 10,
  cl = NULL,
  verbose = TRUE,
  ...
)

# S3 method for class 'CellGraphAssay'
LoadCellGraphs(
  object,
  cells = colnames(object),
  load_as = c("bipartite", "Anode", "linegraph"),
  add_marker_counts = TRUE,
  add_layouts = FALSE,
  force = FALSE,
  chunk_size = 10,
  cl = NULL,
  verbose = TRUE,
  ...
)

# S3 method for class 'CellGraphAssay5'
LoadCellGraphs(
  object,
  cells = colnames(object),
  load_as = c("bipartite", "Anode", "linegraph"),
  add_marker_counts = TRUE,
  add_layouts = FALSE,
  force = FALSE,
  chunk_size = 10,
  cl = NULL,
  verbose = TRUE,
  ...
)

# S3 method for class 'PNAAssay'
LoadCellGraphs(
  object,
  cells = colnames(object),
  add_marker_counts = TRUE,
  add_layouts = FALSE,
  force = FALSE,
  chunk_size = 10,
  cl = NULL,
  verbose = TRUE,
  ...
)

# S3 method for class 'PNAAssay5'
LoadCellGraphs(
  object,
  cells = colnames(object),
  add_marker_counts = TRUE,
  add_layouts = FALSE,
  force = FALSE,
  chunk_size = 10,
  cl = NULL,
  verbose = TRUE,
  ...
)

# S3 method for class 'Seurat'
LoadCellGraphs(
  object,
  assay = NULL,
  cells = colnames(object),
  load_as = c("bipartite", "Anode", "linegraph"),
  add_marker_counts = TRUE,
  add_layouts = FALSE,
  force = FALSE,
  chunk_size = 10,
  cl = NULL,
  verbose = TRUE,
  ...
)

Arguments

object: A Seurat object or an CellGraphAssay object
...: Parameters passed to other methods
cells: A character vector of cell names to load CellGraphs for
load_as: Choose how the cell graph should be represented. This option has no effect when data_type = "PNA" (see details below)
add_marker_counts: Should marker counts be added to the CellGraph objects? Note that this option is ignored for the data.frame method when data_type = "PNA".
data_type: One of "PNA" or "MPX"
chunk_size: Length of chunks used to load CellGraphs from edge list.
verbose: Print messages
add_layouts: Load layouts from the PXL file if available.
force: Force load graph(s) if they are already loaded
cl: A cluster object created by makeCluster, or an integer to indicate number of child-processes (integer values are ignored on Windows) for parallel evaluations. See Details on performance in the documentation for pbapply. The default is NULL, which means that no parallelization is used.
assay: Assay name

Value

An object with a list of CellGraph objects

Details

LoadCellGraphs will only work if the PXL file path(s) stored in the object are valid. You can check the PXL file paths stored in the object with the FSMap method. If the PXL file(s) are invalid, an error will be thrown. Visit RestorePaths for more information on how to update the PXL file paths.

Graph representations

Note that PNA component graphs are always loaded as bipartite graphs.

For MPX data, one of the following graph representations can be used by specifying the load_as parameter:

In the bipartite graph, edges can only go from a upia to a upib. The bipartite graph is first collapsed from a multigraph to a simple graph by aggregating marker counts for each upia/upib combination. For visualization of marker counts on the graph, it is often convenient to project values on the nodes; however, the marker counts are not available in the nodes in the bipartite graph. To circumvent this issue, node counts are calculated by aggregating its edge counts. This means that the total marker count will be inflated.

In the A node projected-graph, pairs of upias that share a upib are connected with an edge. The (upia) node counts are calculated by aggregating counts across all edges associated with the A nodes. The number of nodes in the A node projected-graph is substantially lower than the bipartite graph.

Starting with a bipartite graph, a node is placed on each edge. Then, an edge is drawn between each pair of adjacent nodes. The number of nodes in the linegraph is equivalent to the number of edges in the bipartite graph. However, the number of edges is substantially larger. One major advantage with the linegraph is that the node counts represent the actual raw data, which is not the case for the bipartite graph and the A node projected-graph. On the down side, linegraphs are much larger and tends to slow down layout computations and visualizations.

Examples

# Read from PNAAssay
pxl_file <- minimal_pna_pxl_file()
seur_obj <- ReadPNA_Seurat(pxl_file)
#> ✔ Created a <Seurat> object with 5 cells and 158 targeted surface proteins
pna_assay <- LoadCellGraphs(seur_obj[["PNA"]], cells = "0a45497c6bfbfb22")
#> ℹ Fetching edgelists for 1 cells 
#> → Creating <CellGraph> objects
#> → Fetching marker counts
#> → Adding marker counts to <CellGraph> object(s)
#> ✔ Successfully loaded 1 <CellGraph> object(s).
CellGraphs(pna_assay)[["0a45497c6bfbfb22"]]
#> A CellGraph object containing a bipartite graph with 43543 nodes and 97014 edges
#> Number of markers:  149