Skip to contents

Calculate HVG per batch

rs_sc_hvg_batch_aware.Rd

Batch-aware highly variable gene detection. Calculates HVG statistics separately for each batch, allowing for downstream selection strategies such as union of top genes per batch.

Usage

rs_sc_hvg_batch_aware(
  f_path_gene,
  hvg_method,
  cell_indices,
  batch_labels,
  loess_span,
  clip_max,
  streaming,
  verbose
)

Arguments

f_path_gene

String. Path to the counts_genes.bin file.

hvg_method

String. Which HVG detection method to use. Currently only "vst" is implemented for batch-aware mode.

cell_indices

Integer positions (0-indexed!) that defines the cells to keep.

batch_labels

Integer vector (0-indexed!) defining batch membership for each cell. Must be same length as cell_indices.

loess_span

Numeric. The span parameter for the loess function.

clip_max

Optional clipping number. Defaults to sqrt(no_cells) per batch if not provided.

streaming

Boolean. Shall the genes be streamed in to reduce memory pressure.

verbose

Boolean. Controls verbosity of the function.

Value

A list with HVG statistics concatenated across all batches:

  • mean - The average expression of each gene in each batch.

  • var - The variance of each gene in each batch.

  • var_exp - The expected variance of each gene in each batch.

  • var_std - The standardised variance of each gene in each batch.

  • batch - Batch index for each gene (length = n_genes * n_batches).

  • gene_idx - Gene index for each entry (0-indexed, length = n_genes * n_batches).