Preprocessing steps for Liger dimensionality reduction

Split merged object into multiple sce objects and extract sparse matrices:

liger_preprocess(
  sce,
  k,
  unique_id_var = "manifest",
  take_gene_union = F,
  remove.missing = T,
  num_genes = 2000,
  combine = "union",
  capitalize = F,
  use_cols = T,
  num_cores = 1,
  ...
)

Arguments

sce

SingleCellExperiment object or merged objects

k

Inner dimension of factorization (number of factors).

unique_id_var

the colData variable identifying unique samples. Default is "manifest".

Make a Liger object:

take_gene_union

Whether to fill out raw.data matrices with union of genes across all datasets (filling in 0 for missing data) (requires make.sparse=T) (default FALSE).

remove.missing

Whether to remove cells not expressing any measured genes, and genes not expressed in any cells (if take.gene.union = T, removes only genes not expressed in any dataset) (default TRUE).

Select informative genes:

num_genes

Number of genes to find for each dataset. Set to 3000 as default.

combine

How to combine variable genes across experiments. Either "union" or "intersect". (default "union")

capitalize

Capitalize gene names to match homologous genes (ie. across species) (default FALSE) Scale genes by root-mean-square across cells:

Remove cells/genes with no expression across any genes/cells:

use_cols

Treat each column as a cell (default TRUE)

num_cores

Number of cores used on user's machine to run function. Default is 1.

...

Additional arguments.

Value

liger preprocessed object.

See also

Other Data integration: integrate_sce(), liger_reduce_dims(), report_integrated_sce()