improve testing for fold change

computational-metabolomics · Jul 1, 2024 · a90eeef · a90eeef
1 parent a4fd613
commit a90eeef
Showing 1 changed file with 98 additions and 61 deletions.
diff --git a/tests/testthat/test-foldchange.R b/tests/testthat/test-foldchange.R
@@ -1,64 +1,101 @@
-# fold_change
-test_that('fold_change',{
-  set.seed('57475')
-
-  # data
-  D = iris_DatasetExperiment()
-
-  # two groups
-  F = filter_smeta(mode='exclude',levels='setosa',factor_name='Species')
-  F = model_apply(F,D)
-
-  D = predicted(F)
-
-  # add column for paired data
-  D$sample_meta$sample_id=c(1:50,1:50)
-  D$data[1:50,2] = NA
-  D$data[1:25,3] = NA
-
-  # unpaired
-  FF = fold_change(factor_name='Species',method="geometric",control_group='versicolor')
-  FF = model_apply(FF,D)
-  m=exp(mean(log(D$data[D$sample_meta$Species=='virginica',1]))) / exp(mean(log((D$data[D$sample_meta$Species=='versicolor',1]))))
-  expect_equal(FF$fold_change[1,1],m,tolerance=0.00001)
-  expect_true(is.na(FF$fold_change[2,1]))
-  m=exp(mean(log(na.exclude(D$data[D$sample_meta$Species=='virginica',3])))) / exp(mean(log(na.exclude(D$data[D$sample_meta$Species=='versicolor',3]))))
-  expect_equal(FF$fold_change[3,1],m,tolerance=0.00001)
-
-  FF$method = 'median'
-  FF = model_apply(FF,D)
-  m = median(D$data[D$sample_meta$Species=='virginica',1]) / median(D$data[D$sample_meta$Species=='versicolor',1])
-  expect_equal(FF$fold_change[1,1],m,tolerance=0.00001)
-  expect_true(is.na(FF$fold_change[2,1]))
-  m = median(D$data[D$sample_meta$Species=='virginica',3],na.rm = TRUE) / median(D$data[D$sample_meta$Species=='versicolor',3],na.rm = TRUE)
-  expect_equal(FF$fold_change[3,1],m,tolerance=0.00001)
-
-  FF$method = 'mean'
-  FF = model_apply(FF,D)
-  m = mean(D$data[D$sample_meta$Species=='virginica',1]) / mean(D$data[D$sample_meta$Species=='versicolor',1])
-  expect_equal(FF$fold_change[1,1],m,tolerance=0.00001)
-  expect_true(is.na(FF$fold_change[2,1]))
-  m = mean(D$data[D$sample_meta$Species=='virginica',3],na.rm=TRUE) / mean(D$data[D$sample_meta$Species=='versicolor',3],na.rm=TRUE)
-  expect_equal(FF$fold_change[3,1],m,tolerance=0.00001)
-
-  # paired
-  FF = fold_change(factor_name='Species',method="geometric",paired=TRUE,sample_name = 'sample_id')
-  FF = model_apply(FF,D)
-  m=exp(mean(log(D$data[D$sample_meta$Species=='virginica',1])-log(D$data[D$sample_meta$Species=='versicolor',1])))
-  expect_equal(FF$fold_change[1,1],m,tolerance=0.00001)
-  expect_true(is.na(FF$fold_change[2,1]))
-
-  FF$method = 'median'
-  FF = model_apply(FF,D)
-  m = median(D$data[D$sample_meta$Species=='virginica',1] / D$data[D$sample_meta$Species=='versicolor',1])
-  expect_equal(FF$fold_change[1,1],m,tolerance=0.00001)
-  expect_true(is.na(FF$fold_change[2,1]))
-
-  FF$method = 'mean'
-  FF = model_apply(FF,D)
-  m = mean(D$data[D$sample_meta$Species=='virginica',1] / D$data[D$sample_meta$Species=='versicolor',1])
-  expect_equal(FF$fold_change[1,1],m,tolerance=0.00001)
-  expect_true(is.na(FF$fold_change[2,1]))
 
 
+test_that('fold_change unpaired',{
+    set.seed('57475')
+
+    # data
+    D = iris_DatasetExperiment()
+
+    # add some missing values
+    D$data[1:50,2] = NA
+    D$data[1:25,3] = NA
+
+    # unpaired
+    FF = fold_change(factor_name='Species',method="geometric",control_group='versicolor')
+    FF = model_apply(FF,D)
+    # check some fold changes
+    m=exp(mean(log(na.exclude(D$data[D$sample_meta$Species=='virginica',1])))) / exp(mean(log(na.exclude(D$data[D$sample_meta$Species=='versicolor',1]))))
+    expect_equal(FF$fold_change$`virginica/versicolor`[1],m,tolerance=0.00001)
+
+    m=exp(mean(log(na.exclude(D$data[D$sample_meta$Species=='setosa',4])))) / exp(mean(log(na.exclude(D$data[D$sample_meta$Species=='versicolor',4]))))
+    expect_equal(FF$fold_change$`setosa/versicolor`[4],m,tolerance=0.00001)
+
+    m=exp(mean(log(na.exclude(D$data[D$sample_meta$Species=='virginica',3])))) / exp(mean(log(na.exclude(D$data[D$sample_meta$Species=='setosa',3]))))
+    expect_equal(FF$fold_change$`virginica/setosa`[3],m,tolerance=0.00001)
+
+    # check some NA
+    expect_true(is.na(FF$fold_change$`virginica/setosa`[2]))
+    expect_true(is.na(FF$fold_change$`setosa/versicolor`[2]))
+
+    FF$method = 'median'
+    FF = model_apply(FF,D)
+    # check some fold changes
+    m = median(D$data[D$sample_meta$Species=='virginica',1],na.rm = TRUE) / median(D$data[D$sample_meta$Species=='versicolor',1],na.rm = TRUE)
+    expect_equal(FF$fold_change$`virginica/versicolor`[1],m,tolerance=0.00001)
+
+    m = median(D$data[D$sample_meta$Species=='setosa',4],na.rm = TRUE) / median(D$data[D$sample_meta$Species=='versicolor',4],na.rm = TRUE)
+    expect_equal(FF$fold_change$`setosa/versicolor`[4],m,tolerance=0.00001)
+
+    m = median(D$data[D$sample_meta$Species=='virginica',3],na.rm = TRUE) / median(D$data[D$sample_meta$Species=='setosa',3],na.rm = TRUE)
+    expect_equal(FF$fold_change$`virginica/setosa`[3],m,tolerance=0.00001)
+
+    # check some NA
+    expect_true(is.na(FF$fold_change$`virginica/setosa`[2]))
+    expect_true(is.na(FF$fold_change$`setosa/versicolor`[2]))
+
+    FF$method = 'mean'
+    FF = model_apply(FF,D)
+    m = mean(D$data[D$sample_meta$Species=='virginica',1],na.rm = TRUE) / mean(D$data[D$sample_meta$Species=='versicolor',1],na.rm = TRUE)
+    expect_equal(FF$fold_change$`virginica/versicolor`[1],m,tolerance=0.00001)
+
+    m = mean(D$data[D$sample_meta$Species=='setosa',4],na.rm = TRUE) / mean(D$data[D$sample_meta$Species=='versicolor',4],na.rm = TRUE)
+    expect_equal(FF$fold_change$`setosa/versicolor`[4],m,tolerance=0.00001)
+
+    m = mean(D$data[D$sample_meta$Species=='virginica',3],na.rm = TRUE) / mean(D$data[D$sample_meta$Species=='setosa',3],na.rm = TRUE)
+    expect_equal(FF$fold_change$`virginica/setosa`[3],m,tolerance=0.00001)
+
+    # check some NA
+    expect_true(is.na(FF$fold_change$`virginica/setosa`[2]))
+    expect_true(is.na(FF$fold_change$`setosa/versicolor`[2]))
+
+
 })
+
+test_that('fold_change paired',{
+    set.seed('57475')
+
+    # data
+    D = iris_DatasetExperiment()
+
+    # two groups
+    F = filter_smeta(mode='exclude',levels='setosa',factor_name='Species')
+    F = model_apply(F,D)
+
+    D = predicted(F)
+
+    # add column for paired data
+    D$sample_meta$sample_id=c(1:50,1:50)
+    D$data[1:50,2] = NA
+    D$data[1:25,3] = NA
+
+    # paired
+    FF = fold_change(factor_name='Species',method="geometric",paired=TRUE,sample_name = 'sample_id',control_group = 'versicolor')
+    FF = model_apply(FF,D)
+    m=exp(mean(log(D$data[D$sample_meta$Species=='virginica',1])-log(D$data[D$sample_meta$Species=='versicolor',1])))
+    expect_equal(FF$fold_change[1,1],m,tolerance=0.00001)
+    expect_true(is.na(FF$fold_change[2,1]))
+
+    FF$method = 'median'
+    FF = model_apply(FF,D)
+    m = median(D$data[D$sample_meta$Species=='virginica',1] / D$data[D$sample_meta$Species=='versicolor',1])
+    expect_equal(FF$fold_change[1,1],m,tolerance=0.00001)
+    expect_true(is.na(FF$fold_change[2,1]))
+
+    FF$method = 'mean'
+    FF = model_apply(FF,D)
+    m = mean(D$data[D$sample_meta$Species=='virginica',1] / D$data[D$sample_meta$Species=='versicolor',1])
+    expect_equal(FF$fold_change[1,1],m,tolerance=0.00001)
+    expect_true(is.na(FF$fold_change[2,1]))
+
+
+})