Fit biofluid-stratified rCLR centering — fit_biofluid_stratified

Selects rCLR centering features separately within each training biofluid class. The held-out score is the signed sum of panel log-ratios centered by the denominator selected for the sample's biofluid class.

Usage

fit_biofluid_stratified_rclr(
  train_expr,
  train_meta,
  panel,
  biofluid_col = "biofluid",
  pseudocount = NULL,
  exclude_features = .bfa_default_blood_cell_mirnas(),
  trim_upper = 0.1,
  trim_lower = 0.05,
  min_centering_size = 8L
)

Arguments

train_expr: Numeric matrix, samples x features.
train_meta: Data frame aligned to `train_expr`.
panel: Character vector of panel features.
biofluid_col: Column in `train_meta` with canonical or raw biofluid.
pseudocount: Optional additive pseudocount before log transform.
exclude_features: Features excluded from rCLR denominator selection.
trim_upper, trim_lower: Feature trimming fractions for rCLR denominator selection.
min_centering_size: Minimum denominator size before fallback.

Value

A `biofluid_stratified_rclr_fit`.