Score REO-UCell within-sample rank signature — score_reo

Scores each row of X independently with the frozen signature learned by fit_reo_ucell. For one sample, features in the frozen training universe that are present in X are ranked in decreasing abundance using rank(-x, ties.method = "average"). For a signature $S$ of size $n$, ranks are capped at maxRank_eff + 1, where maxRank_eff = min(model$maxRank, number of present universe features). The UCell-style score is $$1 - \{ \sum_{f \in S} \min(r_f, maxRank_eff + 1) - n(n+1)/2 \} / (n \times maxRank_eff).$$ If no up or down signature features are present, that direction contributes 0. With use_down = TRUE, the returned score is UCell(up) - UCell(down); otherwise it is UCell(up).

Because both the ranking universe and maxRank_eff depend on which model-universe features are present in X, scores are only comparable across specimens evaluated on the same feature panel; score each cohort in a single comparison with one consistent set of features.

Usage

score_reo_ucell(model, X, meta = NULL)

Arguments

model: A reo_ucell_model object returned by fit_reo_ucell.
X: Numeric matrix (samples $\times$ features) of non-negative abundance values. Columns must be named feature ids.
meta: Optional per-sample metadata. Accepted for interface uniformity and ignored by this method.

Value

Numeric vector of one finite score per row of X. Scores use only each row's own values plus the frozen model, with no scored-batch centering, quantiles, or renormalization.

References

Andreatta M, Carmona SJ. (2021) UCell: Robust and scalable single-cell gene signature scoring. Computational and Structural Biotechnology Journal 19: 3796-3798.

Examples

if (FALSE) { # \dontrun{
set.seed(1)
X <- matrix(stats::rgamma(40 * 30, shape = 2), nrow = 40,
            dimnames = list(NULL, paste0("miR-", seq_len(30))))
y <- rep(c(0, 1), each = 20)
X[y == 1, 1:5] <- X[y == 1, 1:5] + 5
model <- fit_reo_ucell(X, y)
score_reo_ucell(model, X[1, , drop = FALSE])
} # }