SCTransform: Use regularized negative binomial regression to normalize UMI...

In lambdamoses/SeuratBasics: Slimmed down version of Seurat

Description

This function calls sctransform::vst. The sctransform package is available at https://github.com/ChristophH/sctransform. Use this function as an alternative to the NormalizeData, FindVariableFeatures, ScaleData workflow. Results are saved in a new assay (named SCT by default) with counts being (corrected) counts, data being log1p(counts), scale.data being pearson residuals; sctransform::vst intermediate results are saved in misc slot of new assay.

Usage

SCTransform(
  object,
  assay = "RNA",
  new.assay.name = "SCT",
  do.correct.umi = TRUE,
  ncells = NULL,
  variable.features.n = 3000,
  variable.features.rv.th = 1.3,
  vars.to.regress = NULL,
  do.scale = FALSE,
  do.center = TRUE,
  clip.range = c(-sqrt(x = ncol(x = object[[assay]])/30), sqrt(x = ncol(x =
    object[[assay]])/30)),
  conserve.memory = FALSE,
  return.only.var.genes = TRUE,
  seed.use = 1448145,
  verbose = TRUE,
  ...
)

Arguments

object	A seurat object
assay	Name of assay to pull the count data from; default is 'RNA'
new.assay.name	Name for the new assay containing the normalized data
do.correct.umi	Place corrected UMI matrix in assay counts slot; default is TRUE
ncells	Number of subsampling cells used to build NB regression; default is NULL
variable.features.n	Use this many features as variable features after ranking by residual variance; default is 3000
variable.features.rv.th	Instead of setting a fixed number of variable features, use this residual variance cutoff; this is only used when variable.features.n is set to NULL; default is 1.3
vars.to.regress	Variables to regress out in a second non-regularized linear regression. For example, percent.mito. Default is NULL
do.scale	Whether to scale residuals to have unit variance; default is FALSE
do.center	Whether to center residuals to have mean zero; default is TRUE
clip.range	Range to clip the residuals to; default is c(-sqrt(n/30), sqrt(n/30)), where n is the number of cells
conserve.memory	If set to TRUE the residual matrix for all genes is never created in full; useful for large data sets, but will take longer to run; this will also set return.only.var.genes to TRUE; default is FALSE
return.only.var.genes	If set to TRUE the scale.data matrices in output assay are subset to contain only the variable genes; default is TRUE
seed.use	Set a random seed. By default, sets the seed to 1448145. Setting NULL will not set a seed.
verbose	Whether to print messages and progress bars
...	Additional parameters passed to sctransform::vst

Value

Returns a Seurat object with a new assay (named SCT by default) with counts being (corrected) counts, data being log1p(counts), scale.data being pearson residuals; sctransform::vst intermediate results are saved in misc slot of the new assay.

See Also

correct_counts get_residuals

Examples

SCTransform(object = pbmc_small)

lambdamoses/SeuratBasics documentation built on May 6, 2020, 9:32 a.m.