Fit kit fixed-effect adjusted ALR. — fit_kit_fe_adjusted

Per feature, log-CPM values are adjusted by subtracting the training stratum mean and adding the global training mean. ALR pivots are selected from features with the lowest between-kit variance. When kit is unknown, biofluid and then global strata are used.

Usage

fit_kit_fe_adjusted_alr(
  expr_matrix,
  sample_meta,
  kit_label_col,
  biofluid_col = "biofluid",
  n_pivots = 6L,
  min_samples_per_stratum = 5L,
  pseudocount = 0.5,
  exclude_features = c("hsa-miR-451a", "hsa-miR-16-5p", "hsa-miR-486-5p",
    "hsa-miR-144-3p", "hsa-miR-223-3p")
)

Arguments

expr_matrix: Numeric matrix, samples x features, counts or non-negative abundance values.
sample_meta: data.frame with one row per sample.
kit_label_col: Column in `sample_meta` containing library-kit family. Missing/unknown kit values fall back to `biofluid_col`, then global.
biofluid_col: Optional fallback column. Default "biofluid".
n_pivots: Number of low-variance pivot features to select.
min_samples_per_stratum: Minimum training samples needed for a stratum-specific centering set.
pseudocount: Additive pseudocount before log-CPM transformation.
exclude_features: Features excluded from denominator selection.

Value

Fit object consumed by `score_kit_fe_adjusted_alr()`.