Subtract reference group median from expression data (relative expression)

This function converts expression into relative (differential) expression by subtracting the per-gene median of a reference group from every sample. The distributed _diff clocks are relative models trained on reference-centred expression, so this centring must always be applied before predicting with them.

Usage

control_subtraction(
  eset,
  column_name = NULL,
  control_label = NULL,
  verbose = TRUE
)

Arguments

eset: An ExpressionSet object containing expression data.
column_name: Character string specifying the column name in phenoData that contains the group labels. Default NULL (centre on all samples).
control_label: Character string specifying the label for control samples. Default NULL (centre on all samples).
verbose: Logical indicating whether to print progress messages and create density plots. Default is TRUE.

Value

An ExpressionSet object with reference-centred expression data.

Details

If no reference group is specified (both column_name and control_label are NULL), or no matching control samples are found, all samples are used as the reference (per-gene overall median). This matches the default behaviour of the TACO reference application, which always centres and defaults the reference group to all samples.

Examples

# Load example data and create ExpressionSet
expr_data <- load_example_expression_data()
meta_data <- load_example_metadata()
eset <- make_ExpressionSet(expr_data, meta_data)
#> ✓ ExpressionSet created successfully
#>   - Number of genes: 57010 
#>   - Number of samples: 24 


# Subtract control group (assuming 'Group' column has 'Control' label)
control_eset <- control_subtraction(eset, "Group", "Control", verbose = TRUE)
#> ✓ No control samples found for label 'Control'. Centring on all samples (overall median).
#> Warning: NaNs produced