Score kit-stratified rCLR.

Usage

score_kit_stratified_rclr(
  expr_matrix,
  sample_meta,
  panel_features,
  kit_label_col,
  biofluid_col = "biofluid",
  fit = NULL,
  feature_weights = NULL,
  pseudocount = NULL,
  ...
)

Arguments

expr_matrix: Numeric matrix, samples x features.
sample_meta: data.frame with one row per sample and `kit_label_col`.
panel_features: Fold-specific panel features, typically top 20.
kit_label_col: Column containing kit family. NA, "unknown", "NR", and non-NGS sentinels fall back to biofluid, then global centering.
biofluid_col: Optional fallback stratum column.
fit: Optional object from `fit_kit_stratified_rclr()`. If omitted, the fit is estimated from `expr_matrix`, which is convenient for deployment but should not be used for held-out benchmarking.
feature_weights: Optional named numeric signs/weights for panel features; default all +1. This lets training folds orient features so higher scores are more case-like.
pseudocount: Optional additive pseudocount before log transform.
...: Additional arguments passed to the fitting helper when `fit` is not supplied.

Value

Numeric score vector of length nrow(expr_matrix).