fine_coarse_subtype_benchmark: fine and coarse subtype benchmark function

In MarianSchoen/DMC: Comparison of different Deconvolution Models in several scenarios

fine and coarse subtype benchmark function

Description

# basically, there are 3 different things to compare. 1. correlation per algorithm on the deep C 2. correlation per algorithm on the accumulated cursory C 3. special scenario where subtype 1 is in X, and subtype 2 is in Y

Usage

fine_coarse_subtype_benchmark(
  sc.counts,
  sc.pheno,
  algorithm.list,
  subtype.pattern = "subtype",
  cell.type.column = "cell_type",
  sample.name.column = "sample.name",
  n.clusters = c(1, 2, 4, 8),
  verbose = FALSE,
  patient.column = "patient",
  n.bulks = 500,
  repeats = 3
)

Arguments

sc.counts	count matrix, features as rows, scRNA-Seq profiles as columns
sc.pheno	data.frame. scRNA-Seq profiles as rows. Must have 'cell.type.column' and 'sample.name.column'
algorithm.list	List containing a list for each algorithm. Each sublist contains 1) name, 2) function and 3) model
subtype.pattern	character, string by which subtype column is recognized; default "subtype"
cell.type.column	string, column of 'sc.pheno' holding the input cell type labels. Within these entries, the clustering is done.
sample.name.column	string, column of the 'colnames(sc.counts)'
n.clusters	integer vector of clustering depths (number of subclusters created for each cell type), default c(1, 2, 4, 8). This means that in the finest clustering, each celltype will be split in 8 subtypes, in the next step each will be split in 4 subtypes, ...
verbose	logical, should information about the process be printed?
patient.column	string, which column of 'pheno' holds the patient information; optional, default "patient"
n.bulks	numeric > 0, number of bulks to simulate. default 500
repeats	numeric > 0, number of repetitions for each algorithm. default: 3

Details

1. and 2. can be done in one go, 3. needs more prepration.

Value

list containing deconvolution results for different cell type granularities

MarianSchoen/DMC documentation built on Aug. 2, 2022, 3:05 p.m.