Integrate embeddings from the integrated sketched.assay

Source: R/integration.R

The main steps of this procedure are outlined below. For a more detailed description of the methodology, please see Hao, et al Biorxiv 2022: doi:10.1101/2022.02.24.481684

ProjectIntegration(
  object,
  sketched.assay = "sketch",
  assay = "RNA",
  reduction = "integrated_dr",
  features = NULL,
  layers = "data",
  reduction.name = NULL,
  reduction.key = NULL,
  method = c("sketch", "data"),
  ratio = 0.8,
  sketched.layers = NULL,
  seed = 123,
  verbose = TRUE
)

Arguments

object

A Seurat object with all cells for one dataset

sketched.assay

Assay name for sketched-cell expression (default is 'sketch')

assay

Assay name for original expression (default is 'RNA')

reduction

Dimensional reduction name for batch-corrected embeddings in the sketched object (default is 'integrated_dr')

features

Features used for atomic sketch integration

layers

Names of layers for correction.

reduction.name

Name to save new reduction as; defaults to paste0(reduction, '.orig')

reduction.key

Key for new dimensional reduction; defaults to creating one from reduction.name

method

Methods to construct sketch-cell representation for all cells (default is 'sketch'). Can be one of:

  • “sketch”: Use random sketched data slot

  • “data”: Use data slot

ratio

Sketch ratio of data slot when dictionary.method is set to “sketch”; defaults to 0.8

sketched.layers

Names of sketched layers, defaults to all layers of “object[[assay]]”

seed

A positive integer. The seed for the random number generator, defaults to 123.

verbose

Print progress and message

Value

Returns a Seurat object with an integrated dimensional reduction

Details

First learn a atom dictionary representation to reconstruct each cell. Then, using this dictionary representation, reconstruct the embeddings of each cell from the integrated atoms.