Select a subset of informative genes — selectGenes • rliger

This function identifies highly variable genes from each dataset and combines these gene sets (either by union or intersection) for use in downstream analysis. Assuming that gene expression approximately follows a Poisson distribution, this function identifies genes with gene expression variance above a given variance threshold (relative to mean gene expression). Alternatively, we allow selecting a desired number of genes for each dataset by ranking the relative variance, and then take the combination.

Usage

selectGenes(object, thresh = 0.1, nGenes = NULL, alpha = 0.99, ...)

# S3 method for liger
selectGenes(
  object,
  thresh = 0.1,
  nGenes = NULL,
  alpha = 0.99,
  useDatasets = NULL,
  useUnsharedDatasets = NULL,
  unsharedThresh = 0.1,
  combine = c("union", "intersection"),
  chunk = 1000,
  verbose = getOption("ligerVerbose", TRUE),
  var.thresh = thresh,
  alpha.thresh = alpha,
  num.genes = nGenes,
  datasets.use = useDatasets,
  unshared.datasets = useUnsharedDatasets,
  unshared.thresh = unsharedThresh,
  tol = NULL,
  do.plot = NULL,
  cex.use = NULL,
  unshared = NULL,
  ...
)

# S3 method for Seurat
selectGenes(
  object,
  thresh = 0.1,
  nGenes = NULL,
  alpha = 0.99,
  useDatasets = NULL,
  layer = "ligerNormData",
  assay = NULL,
  datasetVar = "orig.ident",
  combine = c("union", "intersection"),
  verbose = getOption("ligerVerbose", TRUE),
  ...
)

Arguments

object: A liger, ligerDataset or Seurat object, with normalized data available (no scale factor multipled nor log transformed).
thresh: Variance threshold used to identify variable genes. Higher threshold results in fewer selected genes. Liger and Seurat S3 methods accept a single value or a vector with specific threshold for each dataset in useDatasets.* Default 0.1.
nGenes: Number of genes to find for each dataset. By setting this, we optimize the threshold used for each dataset so that we get nGenes selected features for each dataset. Accepts single value or a vector for dataset specific setting matching useDataset.* Default NULL does not optimize.
alpha: Alpha threshold. Controls upper bound for expected mean gene expression. Lower threshold means higher upper bound. Default 0.99.
...: Arguments passed to other methods.
useDatasets: A character vector of the names, a numeric or logical vector of the index of the datasets to use for shared variable feature selection. Default NULL uses all datasets.
useUnsharedDatasets: A character vector of the names, a numeric or logical vector of the index of the datasets to use for finding unshared variable features. Default NULL does not attempt to find unshared features.
unsharedThresh: The same thing as thresh that is applied to test unshared features. A single value for all datasets in useUnsharedDatasets or a vector for dataset-specific setting.* Default 0.1.
combine: How to combine variable genes selected from all datasets. Choose from "union" or "intersection". Default "union".
chunk: Integer. Number of maximum number of cells in each chunk, when gene selection is applied to any HDF5 based dataset. Default 1000.
verbose: Logical. Whether to show information of the progress. Default getOption("ligerVerbose") or TRUE if users have not set.
var.thresh, alpha.thresh, num.genes, datasets.use, unshared.datasets, unshared.thresh: Deprecated. These arguments are renamed and will be removed in the future. Please see function usage for replacement.
tol, do.plot, cex.use, unshared: Deprecated. Gene variability metric is now visualized with separated function plotVarFeatures. Users can now set none-NULL useUnsharedDatasets to select unshared genes, instead of having to switch unshared on.
layer: Where the input normalized counts should be from. Default "ligerNormData". For older Seurat, always retrieve from data slot.
assay: Name of assay to use. Default NULL uses current active assay.
datasetVar: Metadata variable name that stores the dataset source annotation. Default "orig.ident".

Value

Updated object

liger method - Each involved dataset stored in ligerDataset is updated with its featureMeta slot and varUnsharedFeatures slot (if requested with useUnsharedDatasets), while varFeatures(object) will be updated with the final combined gene set.
Seurat method - Final selection will be updated at Seurat::VariableFeatures(object). Per-dataset information is stored in the meta.features slot of the chosen Assay.

Examples

pbmc <- normalize(pbmc)
#> ℹ Normalizing datasets "ctrl"
#> ℹ Normalizing datasets "stim"
#> ✔ Normalizing datasets "stim" ... done
#> 
#> ℹ Normalizing datasets "ctrl"

#> ✔ Normalizing datasets "ctrl" ... done
#> 
# Select basing on thresholding the relative variance
pbmc <- selectGenes(pbmc, thresh = .1)
#> ℹ Selecting variable features for dataset "ctrl"
#> ✔ ... 168 features selected out of 249 shared features.
#> ℹ Selecting variable features for dataset "stim"
#> ✔ ... 166 features selected out of 249 shared features.
#> ✔ Finally 173 shared variable features are selected.
# Select specified number for each dataset
pbmc <- selectGenes(pbmc, nGenes = c(60, 60))
#> ℹ Selecting variable features for dataset "ctrl"
#> ✔ ... 60 features selected out of 249 shared features.
#> ℹ Selecting variable features for dataset "stim"
#> ✔ ... 60 features selected out of 249 shared features.
#> ✔ Finally 80 shared variable features are selected.