Test per-signature exposure differences between sample groups

The differential-mining half of the embed face (musicatk-style workflow): for each signature, a two-sample Wilcoxon rank-sum test compares the exposures across two groups of samples, with Benjamini-Hochberg correction across the signature family (the same `p.adjust` semantics as ms_test_presence). More than two groups are rejected – pairwise splits are the caller's design decision, not an implicit default.

Usage

ms_exposure_test(exposures, groups, alpha = 0.05)

Arguments

exposures

A k x n exposure matrix (rows = signatures), or an [MsSignature].

groups

A character/factor vector of length n with exactly two levels; the sample order must match the exposure columns.

alpha

The family-wise alpha for `pass_bh` (frozen default 0.05, the presence-test convention).

Value

A data.frame: signature, median_group1, median_group2, statistic, p_raw, p_bh, pass_bh.

References

musicatk (Bioconductor) exposure-differentiation workflow; BH correction semantics per ms_test_presence.