Compute composite scores on test set based on previous cell groups

This function simulates cell subgroups compositions from deconvolution results in the test deconvolution data, and calculates composite scores using provided cell groups and features in order to replicate cell groups coming from a training set into an independent set. The composite scores represent summarized information from cell subgroup profiles.

Usage

compute.test.set(deconv_res, cell_groups, features, deconvolution_test)

Arguments

deconv_res

A list containing deconvolution results, including subgroup compositions (list of data frames or matrices).

cell_groups

A list with three elements:

• cell groups scores

• composition: A named list of character vectors where each element represents cells belonging to a specific group.

• loadings: A corresponding list of numeric vectors (loadings) for each cell group.

features

A character vector of feature names to select relevant cell groups. An error is raised if none of them match a cell group name.

deconvolution_test

A data frame or matrix of deconvolution features for the test set, with cells as columns and samples as rows.

Value

A data frame where each column corresponds to a composite score calculated for each feature group in the test set. If no composite scores can be computed due to zero variance, returns an empty data frame with a printed message.

Details

The function first simulates cell subgroups by computing median values across specified iterations and joins them with the original test deconvolution data. Then it extracts the relevant cells for each feature and calculates composite scores. If the cell groups were built with batch correction, the test samples are centred on their own means (as each training cohort was), so deconvolution_test should contain a single cohort.