mix_samples: Mix samples for loss-function learning DTD

In MarianSchoen/DTD: Digital Tissue Deconvolution

Mix samples for loss-function learning DTD

Description

'mix_samples' takes a gene expresssion 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 "n.samples" samples from "expr.data", and averages over them. Using the information stored in pheno, it can get the quantities per cell in each mixture. Notice, in the mixtures, the frequency of a cell type is reflected by their occurrence in 'pheno'. A cell type that is uncommon, can not occur frequently in the mixtures. In such a case, consider using mix_samples_with_jitter

Usage

mix_samples(
  expr.data,
  pheno,
  included.in.X,
  n.samples = 1000,
  n.per.mixture = 100,
  verbose = FALSE,
  normalize.to.count = TRUE
)

Arguments

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)'
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. Notice, every profile of 'expr.data' might be included in the mixture. But the quantity matrix only reports quantity information for the cell types in 'included.in.X'. This means, that the sum per mixture in the 'quantities' matrix must not add up to 1.
n.samples	integer above 0, numbers of samples to be drawn
n.per.mixture	integer above 0, below ncol(expr.data), how many samples should be included per mixutre
verbose	logical, should information be printed to console?
normalize.to.count	logical, normalize each mixture?

Value

list with two entries: 'mixtures' and 'quantities'.

Examples

library(DTD)
random.data <- generate_random_data(
      n.types = 10,
      n.samples.per.type = 150,
      n.features = 250,
      sample.type = "Cell",
      feature.type = "gene"
      )

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

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

# extract reference matrix X
# First, decide which cells should be deconvoluted.
# Notice, in the mixtures there can be more cells than in the reference matrix.
include.in.X <- paste0("Type", 2:7)

percentage.of.all.cells <- 0.2
sample.X <- sample_random_X(
      included.in.X = include.in.X,
      pheno = indicator.list,
      expr.data = normalized.data,
      percentage.of.all.cells = percentage.of.all.cells
      )
X.matrix <- sample.X$X.matrix
samples.to.remove <- sample.X$samples.to.remove

# all samples that have been used in the reference matrix, must not be included in
# the test/training set
remaining.mat <- random.data[, -which(colnames(random.data) %in% samples.to.remove)]
train.samples <- sample(
      x = colnames(remaining.mat),
      size = ceiling(ncol(remaining.mat)/2),
      replace = FALSE
      )
test.samples <- colnames(remaining.mat)[which(!colnames(remaining.mat) %in% train.samples)]

train.mat <- remaining.mat[, train.samples]
test.mat <- remaining.mat[, test.samples]

indicator.train <- indicator.list[names(indicator.list) %in% colnames(train.mat)]
training.data <- mix_samples(
      expr.data = train.mat,
      pheno = indicator.train,
      included.in.X = include.in.X,
      n.samples = 500,
      n.per.mixture = 100,
      verbose = FALSE
      )

indicator.test <- indicator.list[names(indicator.list) %in% colnames(test.mat)]
test.data <-  mix_samples(
      expr.data = test.mat,
      pheno = indicator.test,
      included.in.X = include.in.X,
      n.samples = 500,
      n.per.mixture = 100,
      verbose = FALSE
      )
# In order to show, that in our mixtures, the sum over the quantities in a
# mixture might not be 1:
sum.per.mixture <- apply(test.data$quantities, 2, sum)
plot(sum.per.mixture)

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