SuperCell (version 1.0)

supercell_2_Seurat: Super-cells to Seurat object

Description

This function transforms super-cell gene expression and super-cell partition into Seurat object

Usage

supercell_2_Seurat(
  SC.GE,
  SC,
  fields = c(),
  var.genes = NULL,
  do.preproc = TRUE,
  is.log.normalized = TRUE,
  do.center = TRUE,
  do.scale = TRUE,
  N.comp = NULL,
  output.assay.version = "v4"
)

Value

Seurat object

Arguments

SC.GE

gene expression matrix with genes as rows and cells as columns

SC

super-cell (output of SCimplify function)

fields

which fields of SC to use as cell metadata

var.genes

set of genes used as a set of variable features of Seurat (by default is the set of genes used to generate super-cells), ignored if !do.preproc

do.preproc

whether to do prepocessing, including data normalization, scaling, HVG, PCA, nearest neighbors, TRUE by default, change to FALSE to speed up conversion

is.log.normalized

whether SC.GE is log-normalized counts. If yes, then Seurat field data is replaced with counts after normalization (see 'Details' section), ignored if !do.preproc

do.center

whether to center gene expression matrix to compute PCA, ignored if !do.preproc

do.scale

whether to scale gene expression matrix to compute PCA, ignored if !do.preproc

N.comp

number of principal components to use for construction of single-cell kNN network, ignored if !do.preproc

output.assay.version

version of the seurat assay in output, `"v4"` by default, `"v5"` requires Seurat v5 installed.

Details

Since the input of CreateSeuratObject should be unnormalized count matrix (UMIs or TPMs, see CreateSeuratObject). Thus, we manually set field `assays$RNA@data` to SC.GE if is.log.normalized == TRUE. Avoid running NormalizeData for the obtained Seurat object, otherwise this will overwrite field `assays$RNA@data`. If you have run NormalizeData, then make sure to replace `assays$RNA@data` with correct matrix by running `your_seurat@assays$RNA@data <- your_seurat@assays$RNA@counts`.

Since super-cells have different size (consist of different number of single cells), we use sample-weighted algorithms for all possible steps of the downstream analysis, including scaling and dimensionality reduction. Thus, generated Seurat object comes with the results of sample-wighted scaling (available as `your_seurat@assays$[e-mail]` or `your_seurat@assays$RNA@misc[["scale.data.weighted"]]` to reproduce if the first one has been overwritten) and PCA (available as `your_seurat@reductions$pca` or `your_seurat@reductions$pca_weighted` to reproduce if the first one has been overwritten).

Examples

# \donttest{
data(cell_lines)
SC           <- SCimplify(X=cell_lines$GE, gamma = 20)
SC$ident     <- supercell_assign(clusters = cell_lines$meta, supercell_membership = SC$membership)
SC.GE        <- supercell_GE(cell_lines$GE, SC$membership)
m.seurat     <- supercell_2_Seurat(SC.GE = SC.GE, SC = SC, fields = c("ident"))
# }