Skip to contents

Identify HVGs (batch aware)

find_hvg_batch_aware_sc.Rd

This is a helper function to identify highly variable genes in a batch-aware manner. At the moment the implementation has only the VST-based version (known as Seurat v3). The other methods will be implemented in the future. This function will calculate the HVG per given experimental batch and you can choose the way to combine them. The choices are union (of Top x HVG per batch), based on the average variance per batch or only take genes that are amongst the Top X HVG in all batches. Important. The function returns 0-indices for the genes!

Usage

find_hvg_batch_aware_sc(
  object,
  batch_column,
  hvg_no = 2000L,
  gene_comb_method = c("union", "average", "intersection"),
  hvg_params = params_sc_hvg(),
  streaming = FALSE,
  .verbose = TRUE
)

Arguments

object

SingleCells class.

batch_column

String. The column name of the batch column in the obs table.

hvg_no

Integer. Number of highly variable genes to include. Defaults to 2000L.

gene_comb_method

String. One of c("union", "average", "intersection"). The method to combine the HVG across the different batches. Defaults to "union".

hvg_params

List, see params_sc_hvg(). This list contains

  • method - Which method to use. One of c("vst", "meanvarbin", "dispersion")

  • loess_span - The span for the loess function to standardise the variance

  • num_bin - Integer. Not yet implemented.

  • bin_method - String. One of c("equal_width", "equal_freq"). Not implemented yet.

streaming

Boolean. Shall the genes be streamed in. Useful for larger data sets where you wish to avoid loading in the whole data. Defaults to FALSE.

.verbose

Boolean. Controls verbosity and returns run times.

Value

This function will return a list with:

  • hvg_indices - The indices of the batch-aware HVG.

  • batch_hvg_data - data.table with the detailed information of the variance per batch.