Fit kit-orthogonal ILR-like residualization. — fit_kit_orthogonal

The training CLR feature space is used to estimate the dominant kit axis between the two most populated known strata. Scores are computed after removing the projection onto that axis. If fewer than two strata are available, scoring falls back to the global CLR panel sum and the fit status records the limitation.

Usage

fit_kit_orthogonal_ilr(
  expr_matrix,
  sample_meta,
  kit_label_col,
  biofluid_col = "biofluid",
  min_samples_per_stratum = 5L,
  pseudocount = NULL
)

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".
min_samples_per_stratum: Minimum training samples needed for a stratum-specific centering set.
pseudocount: Optional additive pseudocount before log transform.

Value

Fit object consumed by `score_kit_orthogonal_ilr()`.