Within-Sample Pairwise Log-Ratio Features

Computes all pairwise log-ratios between features within each sample. For p features, produces p*(p-1)/2 ratio features. Each ratio is inherently self-normalizing — multiplicative batch effects cancel.

If the input is already on log scale (e.g., log2 expression), the log-ratio is simply the difference: log(A/B) = log(A) - log(B).

Usage

ws_logratio(x, feature_names = NULL)

Arguments

x

A numeric matrix (samples x features). Assumed log-scale.

feature_names

Optional character vector of feature names

Value

Matrix with p*(p-1)/2 columns (pairwise ratios)