tests/testthat/test_density.R

In PerkinElmer/phenoptr: inForm Helper Functions

# Test for density_bands
library(testthat)
library(dplyr)

test_that('density_bands works', {
  # Compute density for the sample data
  values <- density_bands(sample_cell_seg_path(),
    list("CD8+", "CD68+", "FoxP3+"),
    positive="Stroma", negative="Tumor")

  expect_equal(names(values), c("densities", "cells", "distance"))

  densities = values$densities
  expect_equal(range(densities$midpoint), c(-112.5, 212.5))

  dens_summary = densities %>% group_by(phenotype) %>%
    summarize(count=sum(count), pixels=sum(area)*4)

  # Sum of pixels is image size
  expect_equal(dens_summary$pixels, rep(1868*1400, 3))

  # Cell counts should match cell_seg_data
  csd_summary = sample_cell_seg_data %>% group_by(Phenotype) %>%
    summarize(count=n()) %>%
    right_join(dens_summary, by=c(Phenotype='phenotype'))

  expect_equal(nrow(csd_summary), 3)
  expect_equal(csd_summary$count.x, csd_summary$count.y)

  # Now by tissue
  dens_summary = densities %>%
    mutate(tissue = ifelse(midpoint<0, 'Tumor', 'Stroma')) %>%
    group_by(tissue, phenotype) %>%
    summarize(count=sum(count))

  # Cell counts should be close to cell_seg_data
  # There are differences at the tissue boundary...
  csd_summary = sample_cell_seg_data %>%
    group_by(`Tissue Category`, Phenotype) %>%
    summarize(count=n()) %>%
    right_join(dens_summary,
               by=c('Tissue Category'='tissue', Phenotype='phenotype'))

  expect_equal(nrow(csd_summary), 6)
  expect_equal(csd_summary$count.x, csd_summary$count.y, tolerance=4)
})