Compute deconvolution benchmark

Usage

compute.benchmark(
  deconvolution,
  groundtruth,
  cells_extra = NULL,
  corr_type = "spearman",
  scatter = TRUE,
  plot = FALSE,
  pval = 0.05,
  file_name = NULL,
  width = 16,
  height = 8
)

Arguments

deconvolution: The deconvolution matrix output from compute.deconvolution()
groundtruth: A matrix with the cell type proportions (samples as rows and cell types as columns). Cell types names should correspond to the ones on the deconvolution matrix.
cells_extra: A string specifying the cells names to consider and that are not including in the nomenclature of multideconv (see Readme)
corr_type: Secifies the type of correlations to compute ('spearman' or 'pearson').
scatter: Boolean value to specify if scatter plots should be returned.
plot: Boolean value to whether save or not the plot of the benchmark in the Results/ directory.
pval: A numeric value with the pvalue to use for selecting significant features.
file_name: A string specifying the name of the plot saved in Results/
width: A numeric value with the width for the returned plot.
height: A numeric value with the height for the returned plot.

Value

A correlation matrix between the cell type deconvolution combinations and the real cell proportions, with an "average" row (mean over cell types) used to order the combinations. When deconvolution contains subgroups (e.g. B.cells_Subgroup.1), columns are Subgroup.1, Subgroup.2, ... (the i-th subgroup of each cell type) and no average is computed, because a column then holds unrelated features.

Examples


data("deconvolution")
data("cells_groundtruth")

corr_matrix = compute.benchmark(deconvolution, cells_groundtruth, cells_extra = "Myeloid.cells",
                                corr_type = "spearman", scatter = FALSE)
#> No id variables; using all as measure variables
#> No id variables; using all as measure variables