Tree probability for one cell

Compute the probability that a single cell belongs to a given cell type by combining per-node scores along a tree path, applying a complexity-based weight, and an optional negative marker penalty.

Usage

tree_prob(
  tree,
  expr_mat,
  cell_i,
  combine = c("mean", "product"),
  neg_markers = character(0),
  marker_stats = NULL,
  neg_strength = 0.8,
  lambda = 2
)

Arguments

tree

A gating tree as produced by [build_fullcoverage_tree()].

expr_mat

Numeric matrix-like expression object with markers in rows and cells in columns.

cell_i

Integer index of the cell/column to score.

combine

How to combine node-wise scores along the path. One of `"mean"` or `"product"`.

neg_markers

Character vector of negative marker names used for penalty.

marker_stats

A list of marker statistics (including cutoff and scale) produced by [fit_marker_stats()] used to calculate the penalty.

neg_strength

Numeric scalar in [0,1] defining how strongly negative marker expression should penalize the final score. Default is 0.8.

lambda

Numeric scalar representing the "Skepticism Factor." Higher values place a heavier penalty on cell types defined by fewer markers. Default is 0.2.

Value

Numeric scalar probability (typically in [0,1]) or `NA_real_` if no usable node scores are available for the cell.

Examples

tree <- list(
  type = "node", marker = "CD3", cutoff = 1, scale = 0.5,
  left = list(type = "leaf"), right = list(type = "leaf")
)
expr_mat <- rbind(CD3 = c(0.5, 1.5))
tree_prob(tree, expr_mat, cell_i = 2)
#> [1] 0.2436862