mix_samples_with_jitter: Mix samples with jitter

In MarianSchoen/DTD: Digital Tissue Deconvolution

Mix samples with jitter

Description

'mix_samples_with_jitter' takes a expression matrix ('expr.data') and 'pheno' information. It then mixes the samples with known quantities, such that it can be used for loss-function learning digital tissue deconvolution. For a mixture it randomly selects a quantity for each cell type. Then, it randomly selects profiles from 'expr.data' for each cell type, multiplies it with the respective quantity, and averages it to a simulated bulk profile. Notice, in the mixtures, the frequency of a cell type is reflected by the 'prob.each' vector, and not by their occurrence in 'pheno'. Alternatively, there is a 'mix_samples' function, that reflects the underlying composition of the data set: mix_samples

Usage

mix_samples_with_jitter(
  included.in.X,
  prob.each = NA,
  n.samples,
  expr.data,
  pheno,
  verbose = FALSE,
  add.jitter = FALSE,
  chosen.mean = 1,
  chosen.sd = 0.05,
  n.per.mixture = 1,
  normalize.to.count = TRUE
)

Arguments

included.in.X	vector of strings, indicating types that are in the reference matrix. Only those types, and sorted in that order, will be included in the quantity matrix.
prob.each	numeric vector with same length as 'included.in.X.' For each cell type in 'included.in.X', prob.each' holds the expected average quantity in the mixtures.
n.samples	integer above 0, numbers of samples to be drawn
expr.data	numeric matrix, with features as rows and samples as columns
pheno	named vector of strings, with pheno information ('pheno') for each sample in 'expr.data'. names(pheno)' must all be in 'colnames(expr.data)'
verbose	logical, should information be printed to console?
add.jitter	logical, should each mixture be multiplied with a vector of normally distributed numbers? (JITTER)
chosen.mean	float, mean of jitter
chosen.sd	float, standard deviation of jitter
n.per.mixture	integer above 0, below ncol(expr.data), how many samples should be included per mixutre
normalize.to.count	logical, normalize each mixture?

Value

list with two entries. "quantities": matrix (nrow = ncol(expr.data), ncol = n.samples) and "mixtures": matrix (nrow = nrow(expr.data), ncol = n.samples)

Examples

library(DTD)
random.data <- generate_random_data(
      n.types = 10,
      n.samples.per.type = 10,
      n.features = 500,
      sample.type = "Cell",
      feature.type = "gene"
      )

# normalize all samples to the same amount of counts:
random.data <- normalize_to_count(random.data)

# extract indicator list.
# This list contains the type of the sample as value, and the sample name as names
indicator.list <- gsub("^Cell[0-9]*\\.", "", colnames(random.data))
names(indicator.list) <- colnames(random.data)

# First, decide which cells should be deconvoluted.
included.in.X <- c("Type2", "Type3", "Type4", "Type5")

training.data <- mix_samples_with_jitter(
    included.in.X = included.in.X
    , prob.each = c(2,1,1,1)
    , n.samples = 1e3
    , expr.data = random.data
    , pheno = indicator.list
    , add.jitter = TRUE
    , chosen.mean = 1
    , chosen.sd = 0.05
)

# see the effect of "Type2" having higher 'prob.each' entry:
apply(training.data$quantities, 1, mean)

MarianSchoen/DTD documentation built on April 29, 2022, 1:59 p.m.