tests/testthat/test-multiSingscore.R

In singscore: Rank-based single-sample gene set scoring method

library(GSEABase)
context("multiSingscore")

test_that("checkGenesMulti works", {
  #set seed for reproducibility
  set.seed(36)

  #generate geneset collection
  gsl = lapply(1:10, function(x)
    GeneSet(as.character(sample.int(100, 5)), setName = LETTERS[x]))
  gsc = GeneSetCollection(gsl)

  gsc_filt = checkGenesMulti(gsc, as.character(1:20))
  expect_lt(length(gsc_filt), length(gsc))
  expect_equal(all(sapply(gsc_filt, function(x) length(geneIds(x))) < 5), TRUE)
})

test_that("multiSingscore works", {
  #set seed for reproducibility
  set.seed(36)

  #generate geneset collection
  gsl = lapply(1:20, function(x)
    GeneSet(as.character(sample.int(100, 5)), setName = LETTERS[ceiling(x / 2)]))
  gsc_up = gsl[(1:20) %% 2 == 0]
  gsc_dn = gsl[(1:20) %% 2 == 1]

  #generate data
  emat = matrix(rnorm(6 * 80), ncol = 6)
  rownames(emat) = as.character(21:100)
  eranks = rankGenes(emat)

  #with lists of GeneSets
  expect_equal(is.array(multiScore(eranks, gsc_up[1])$Score), TRUE)
  expect_equal(length(multiScore(eranks, gsc_up)), 2)
  expect_equal(ncol(multiScore(eranks, gsc_up)$Score), 6)
  expect_equal(ncol(multiScore(eranks, gsc_up, gsc_dn)$Score), 6)
  expect_equal(ncol(multiScore(eranks, gsc_up, knownDirection = FALSE)$Score), 6)
  
  #with GeneSetCollections
  gsc_up = GeneSetCollection(gsc_up)
  gsc_dn = GeneSetCollection(gsc_dn)
  expect_equal(is.array(multiScore(eranks, gsc_up[1])$Score), TRUE)
  expect_equal(length(multiScore(eranks, gsc_up)), 2)
  expect_equal(ncol(multiScore(eranks, gsc_up)$Score), 6)
  expect_equal(ncol(multiScore(eranks, gsc_up, gsc_dn)$Score), 6)
  expect_equal(ncol(multiScore(eranks, gsc_up, knownDirection = FALSE)$Score), 6)
  
  #with an empty GeneSet in the list
  geneIds(gsc_dn[[3]]) = as.character(1000:1005)
  expect_equal(is.array(multiScore(eranks, gsc_dn[1])$Score), TRUE)
  expect_equal(length(multiScore(eranks, gsc_dn)), 2)
  expect_equal(ncol(multiScore(eranks, gsc_dn)$Score), 6)
  expect_equal(nrow(multiScore(eranks, gsc_dn)$Score), length(gsc_up))
  expect_true(all(is.na(multiScore(eranks, gsc_dn)$Score['C', ])))
  expect_equal(nrow(multiScore(eranks, gsc_up, gsc_dn)$Score), length(gsc_up))
  expect_true(all(is.na(multiScore(eranks, gsc_up, gsc_dn)$Score['C', ])))
})

test_that('Multiscore works for a single sample', {
  set.seed(36)
  
  #generate geneset collection
  gsl = lapply(1:20, function(x)
    GeneSet(as.character(sample.int(100, 5)), setName = LETTERS[ceiling(x / 2)]))
  gsc_up = gsl[(1:20) %% 2 == 0]
  gsc_dn = gsl[(1:20) %% 2 == 1]
  
  #generate data
  emat = matrix(rnorm(6 * 80), ncol = 6)
  rownames(emat) = as.character(21:100)
  
  #single sample measurements given
  eranks = rankGenes(emat[, 1, drop = FALSE])
  expect_length(multiScore(eranks, gsc_up, knownDirection = FALSE), 2)
  
  #single sample selected
  eranks = rankGenes(emat)
  expect_length(multiScore(eranks, gsc_up, subSamples = 1, knownDirection = FALSE), 2)
  
  #single sample with a list of just one geneset
  expect_length(multiScore(eranks, gsc_up[1], subSamples = 1, knownDirection = FALSE), 2)
})

test_that('Multiscore using stable genes evaluates correctly', {
  df = as.data.frame(c(1,2,5,5,5))
  rownames(df) = LETTERS[1:5]
  colnames(df) = 'test'
  gsDn = GSEABase::GeneSet(as.character(c('A')))
  GSEABase::setName(gsDn) = 'Dn'
  gsUp = GSEABase::GeneSet(as.character(c('D', 'E')))
  GSEABase::setName(gsUp) = 'Up'
  gslist = list(gsUp, gsDn)
  
  eranks = rankGenes(df, tiesMethod = 'min', stableGenes = c('A', 'C'))
  expect_equal(as.vector(multiScore(eranks, gslist)$Scores), c(1/6, -1/6))
})