tests/testthat/test-pca_cor_samplevar.R

In insightsengineering/hermes: Preprocessing, analyzing, and reporting of RNA-seq data

# h_pca_var_rsquared ----

test_that("h_pca_var_rsquared works as expected", {
  object <- hermes_data
  pca <- expect_silent(calc_pca(object)$x)
  x <- expect_silent(colData(object)$low_depth_flag)
  r2 <- expect_silent(h_pca_var_rsquared(pca, x))
  expect_numeric(r2)
  expect_true(noNA(r2))
  expect_identical(ncol(pca), length(r2))
})

test_that("h_pca_var_rsquared fails as expected with invalid inputs", {
  se <- get_se()
  object <- HermesData(se)
  x <- expect_silent(colData(se)$low_depth_flag)
  expect_error(h_pca_var_rsquared(se, x))
  expect_error(h_pca_var_rsquared(object, x))
})

test_that("h_pca_var_rsquared returns warning when something is not estimable", {
  pca <- rbind(X = -1, Y = 1)
  x <- c(FALSE, FALSE)
  expect_warning(
    expect_warning(
      h_pca_var_rsquared(pca, x),
      "sample variable is constant and R2 values cannot be calculated"
    ),
    "Partial NA coefficients"
  )
})

# h_pca_df_r2_matrix ----

test_that("h_pca_df_r2_matrix works as expected", {
  object <- hermes_data
  pca <- expect_silent(calc_pca(object)$x)
  df <- expect_silent(as.data.frame(df_cols_to_factor(colData(object))))
  result <- expect_silent(h_pca_df_r2_matrix(pca, df))
  expect_matrix(result)
  expect_identical(nrow(result), ncol(pca))
  expect_lt(ncol(result), ncol(df))
  expect_true(all(colnames(result) %in% colnames(df)))
})

test_that("h_pca_df_r2_matrix fails as expected with invalid settings", {
  object <- hermes_data
  pca <- expect_silent(calc_pca(object)$x)
  df <- expect_silent(as.data.frame(colData(object)))
  expect_error(h_pca_df_r2_matrix(pca, object))
  expect_error(h_pca_var_rsquared(object, df))
})

# correlate-HermesDataPca ----

test_that("correlate method on HermesDataPca works as expected", {
  object <- hermes_data
  df <- expect_silent(as.data.frame(colData(object)))
  pca <- expect_silent(calc_pca(object))
  result <- expect_silent(correlate(pca, object))
  is(result, "HermesDataPcaCor")
  expect_identical(nrow(result), ncol(pca$x))
  expect_lt(ncol(result), ncol(df))
  expect_identical(rownames(result), colnames(pca$x))
  expect_true(all(colnames(result) %in% colnames(df)))
})

test_that("correlate method fails as expected with invalid settings", {
  object <- hermes_data
  pca <- expect_silent(calc_pca(object))
  result <- expect_silent(correlate(pca, object))
  expect_error(correlate(pca, colData(object)))
  expect_error(correlate(pca$x, object))
})

# autoplot-HermesDataPcaCor ----

test_that("autoplot method does not give warnings on HermesDataPcaCor objects", {
  hd <- hermes_data
  object <- hd %>%
    calc_pca() %>%
    correlate(hd)
  expect_s4_class(object, "HermesDataPcaCor")
  result <- expect_silent(autoplot(object))
  expect_s4_class(result, "Heatmap")
})