Prevalence comparison by group (POP framework)

Computes feature-level prevalence across binary group columns and performs pairwise statistical tests:

Unpaired: Fisher's exact test (2x2) per (rank, feature, group pair).
Paired: McNemar's exact binomial test per subject.

Presence per sample is determined by a k-of-n rule: a feature is considered present in a sample if at least pop_k_min of its contributing peptides are positive. Each group_col must have exactly 2 non-missing levels in the data.

Usage

compute_pop(
  x,
  rank_cols,
  group_cols,
  exist_col = "exist",
  pop_k_min = 1L,
  paired = FALSE,
  peptide_library = NULL
)

Arguments

x: A phip_data object or a data.frame/tibble with at least columns sample_id, peptide_id, exist_col, and all group_cols. If paired is set, the paired linking column must also be present. If rank_cols include non-peptide taxa, x must provide peptide_library.
rank_cols: Character vector of rank columns, e.g. c("peptide_id", "species").
group_cols: Character vector of binary grouping columns. Each column must have exactly 2 non-missing levels in the data.
exist_col: Name of the binary presence column (default "exist").
pop_k_min: Integer >= 1; k-of-n POP threshold per sample (default 1).
paired: FALSE (default) or a single string naming the column that links related samples (e.g. "subject_id"). When set, paired McNemar tests are used instead of Fisher.
peptide_library: Optional peptide metadata table with peptide_id and any non-peptide rank columns. If NULL, taken from x$peptide_library.

Value

A data.frame with one row per (rank, feature, group_col) comparison: rank, feature, group_col, group1, group2, n1, N1, prop1, percent1, n2, N2, prop2, percent2, ratio, delta_ratio (unpaired only), p_raw, n_peptides. A view column is prepended when the input carries a view attribute.

Examples

if (FALSE) { # \dontrun{
res <- compute_pop(pd, rank_cols = "species", group_cols = "group")
res
} # }