Skip to contents

Transfer cell scores between matrices using gene weights

Source: R/H_extrapolate.R
transfer_scores.Rd

Given reference (mat_A) and target (mat_B) cell-by-gene matrices with the same set and order of genes (columns), this function optionally quantile normalizes the target to match the reference per feature, standardizes the target using the reference mean and standard deviation, filters small-magnitude gene weights, and computes cell scores via matrix multiplication \(B_{standardized} \\times W\).

Usage

transfer_scores(
  mat_A,
  mat_B,
  gs_ct,
  use_quantile_normalization = TRUE,
  gs_weight_threshold = 0,
  verbose = TRUE
)

Arguments

mat_A

Numeric matrix of shape cells-by-genes (reference). Column order must correspond to genes in gs_ct.

mat_B

Numeric matrix of shape cells-by-genes (target). Must have the same genes (columns) and order as mat_A.

gs_ct

Numeric matrix of gene weights of shape genes-by-K, where K is the number of signatures or cell types. Column names, if present, are used in verbose messages.

use_quantile_normalization

Logical; whether to quantile-normalize mat_B to mat_A per feature before standardization (default TRUE).

gs_weight_threshold

Numeric; absolute threshold used to zero out small gene weights in gs_ct (default 0).

verbose

Logical; whether to print progress messages (default TRUE).

Value

A numeric matrix of cell scores with shape (nrow(mat_B) x ncol(gs_ct)).

Details

  • If use_quantile_normalization = TRUE, mat_B is quantile-normalized to the distribution of mat_A for each gene.

  • Columns of the normalized mat_B are then centered and scaled using the column means and standard deviations computed from mat_A (with a small safeguard for near-zero standard deviations).

  • Gene weights in gs_ct are thresholded by absolute value per column; values with abs(weight) < gs_weight_threshold are set to 0.

  • Final cell scores are obtained as the matrix product of the standardized mat_B with the filtered weight matrix gs_ct.

This function assumes that mat_A, mat_B, and gs_ct are aligned on the same set and ordering of genes (columns of mat_A/mat_B, rows of gs_ct). Callers are responsible for subsetting/reordering if needed.

See also

getTransferCellScores