Integration method functions can be written by anyone to implement any
integration method in Seurat. These methods should expect to take a
v5 assay as input and return a named list of
objects that can be added back to a Seurat object (eg. a
dimensional reduction or cell-level meta data)
Every integration method function should expect the following arguments:
“object”: an Assay5 object
“orig”: dimensional
reduction to correct
“layers”: names of normalized layers in object
“scale.layer”: name(s) of scaled layer(s) in
object
“features”: a vector of features for integration
“groups”: a one-column data frame with the groups for
each cell in object; the column name will be “group”
The documentation for IntegrateLayers() will automatically
link to integration method functions provided by packages in the
search() space. To make an integration method function
discoverable by the documentation, simply add an attribute named
“Seurat.method” to the function with a value of
“integration”
attr(MyIntegrationFunction, which = "Seurat.method") <- "integration"