convertPCsToSCE: Convert corrected PCs to a SingleCellExperiment

View source: R/convertPCsToSCE.R

convertPCsToSCER Documentation

Convert corrected PCs to a SingleCellExperiment

Description

Convert low-dimensional corrected PCs to a SingleCellExperiment containing corrected expression values. This is a low-level function and most users should not need to call it.

Usage

convertPCsToSCE(
  corrected.df,
  pc.info,
  assay.name = "reconstructed",
  dimred.name = "corrected"
)

Arguments

corrected.df

A DataFrame containing a nested matrix of low-dimensional corrected values and a vector of batch identities. Typically produced from reducedMNN.

pc.info

A list containing PCA statistics, in particular a rotation matrix. Typically obtained from the metadata of the output from multiBatchPCA.

assay.name

String specifying the name of the assay to use to store the corrected expression values.

dimred.name

String containing the name fo the reducedDims to store the low-dimensional corrected values. Set to NULL to avoid storing these.

Details

The corrected expression values are obtained by simply taking the crossproduct of the corrected PCs with the rotation matrix. This reverses the original projection to PC space while retaining the effect of the correction. These values are best used for visualization; the low-dimensional corrected coordinates are more efficient for per-cell operations like clustering, while the original uncorrected expression values are safer to interpret for per-gene analyses.

Value

A SingleCellExperiment containing a LowRankMatrix with the corrected per-gene expression values. The colData contains the batch identities, the rowData contains the rotation matrix, and the reducedDims contains the low-dimensional corrected values (if dimred.name is not NULL). All additional metadata from corrected.df and pc.info is stored in metadata.

Author(s)

Aaron Lun

See Also

reducedMNN, to compute corrected.df; and multiBatchPCA, to compute pc.info.

fastMNN, which uses this function to obtain low-rank corrected values.

Examples

B1 <- matrix(rnorm(10000), nrow=50) # Batch 1
B2 <- matrix(rnorm(10000), nrow=50) # Batch 2

# Equivalent to fastMNN().
cB1 <- cosineNorm(B1)
cB2 <- cosineNorm(B2)
pcs <- multiBatchPCA(cB1, cB2)
mnn.out <- reducedMNN(pcs[[1]], pcs[[2]])

sce <- convertPCsToSCE(mnn.out, metadata(pcs))
sce


LTLA/batchelor documentation built on July 10, 2024, 9:09 p.m.