# ---------------------------------------------------------------------------- # Copyright (c) 2017-2023, QIIME 2 development team. # # Distributed under the terms of the Modified BSD License. # # The full license is in the file LICENSE, distributed with this software. # ---------------------------------------------------------------------------- import pandas as pd import pandas.testing as pdt from os import mkdir, listdir from os.path import join import biom import qiime2 from qiime2.plugins import sample_classifier from q2_sample_classifier.visuals import ( _linear_regress, _calculate_baseline_accuracy, _add_sample_size_to_xtick_labels) from q2_sample_classifier.classify import ( scatterplot, confusion_matrix) from q2_sample_classifier.utilities import ( _match_series_or_die, _predict_and_plot) from q2_sample_classifier.tests.test_base_class import \ SampleClassifierTestPluginBase class TestVisuals(SampleClassifierTestPluginBase): md = pd.DataFrame([(1, 'a', 0.11), (1, 'a', 0.12), (1, 'a', 0.13), (2, 'a', 0.19), (2, 'a', 0.18), (2, 'a', 0.21), (1, 'b', 0.14), (1, 'b', 0.13), (1, 'b', 0.14), (2, 'b', 0.26), (2, 'b', 0.27), (2, 'b', 0.29)], columns=['Time', 'Group', 'Value']) def test_linear_regress(self): res = _linear_regress(self.md['Value'], self. md['Time']) self.assertAlmostEqual(res.iloc[0]['Mean squared error'], 1.9413916666) self.assertAlmostEqual(res.iloc[0]['r-value'], 0.86414956372460128) self.assertAlmostEqual(res.iloc[0]['r-squared'], 0.74675446848541871) self.assertAlmostEqual(res.iloc[0]['P-value'], 0.00028880275858705694) def test_calculate_baseline_accuracy(self): accuracy = 0.9 y_test = pd.Series(['a', 'a', 'a', 'b', 'b', 'b'], name="class") classifier_accuracy = _calculate_baseline_accuracy(y_test, accuracy) expected_results = (6, 3, 0.5, 1.8) for i in zip(classifier_accuracy, expected_results): self.assertEqual(i[0], i[1]) class TestHeatmap(SampleClassifierTestPluginBase): def setUp(self): super().setUp() md_vaw = self.get_data_path('vaw.txt') md_vaw = qiime2.Metadata.load(md_vaw) self.md_vaw = md_vaw.get_column('Column') table_vaw = self.get_data_path('vaw.qza') self.table_vaw = qiime2.Artifact.load(table_vaw) imp = pd.read_csv( self.get_data_path('vaw_importance.tsv'), sep='\t', header=0, index_col=0) self.imp = qiime2.Artifact.import_data('FeatureData[Importance]', imp) def test_heatmap_default_feature_count_zero(self): heatmap, table, = sample_classifier.actions.heatmap( self.table_vaw, self.imp, self.md_vaw, group_samples=True, feature_count=0) self.assertEqual(table.view(biom.Table).shape, (5, 2)) def test_heatmap_importance_threshold(self): heatmap, table, = sample_classifier.actions.heatmap( self.table_vaw, self.imp, self.md_vaw, importance_threshold=0.062, group_samples=False, feature_count=0) self.assertEqual(table.view(biom.Table).shape, (3, 6)) def test_heatmap_feature_count(self): heatmap, table, = sample_classifier.actions.heatmap( self.table_vaw, self.imp, self.md_vaw, group_samples=True, feature_count=2) self.assertEqual(table.view(biom.Table).shape, (2, 2)) def test_heatmap_must_group_or_die(self): with self.assertRaisesRegex(ValueError, "metadata are not optional"): heatmap, table, = sample_classifier.actions.heatmap( self.table_vaw, self.imp, sample_metadata=None, group_samples=True) # This class really just checks that these visualizers run without error. Yay. # Also test some internal nuts/bolts but there's not much else we can do. class TestPlottingVisualizers(SampleClassifierTestPluginBase): def setUp(self): super().setUp() self.tmpd = join(self.temp_dir.name, 'viz') mkdir(self.tmpd) self.a = pd.Series(['a', 'a', 'b', 'b', 'c', 'c'], name='site', index=['a1', 'a2', 'b1', 'b2', 'c1', 'c2']) self.a.index.name = 'SampleID' self.bogus = pd.Series(['a', 'a', 'b', 'b', 'c', 'c'], name='site', index=['a1', 'e3', 'f5', 'b2', 'z1', 'c2']) self.bogus.index.name = 'SampleID' self.c = pd.Series( [0, 1, 2, 3], index=['a', 'b', 'c', 'd'], name='peanuts') self.c.index.name = 'SampleID' def test_confusion_matrix(self): b = qiime2.CategoricalMetadataColumn(self.a) confusion_matrix(self.tmpd, self.a, b) def test_confusion_matrix_class_overlap_error(self): b = pd.Series([1, 2, 3, 4, 5, 6], name='site', index=['a1', 'a2', 'b1', 'b2', 'c1', 'c2']) b.index.name = 'id' b = qiime2.NumericMetadataColumn(b) with self.assertRaisesRegex(ValueError, "do not overlap"): confusion_matrix(self.tmpd, self.a, b) def test_confusion_matrix_vmin_too_high(self): b = qiime2.CategoricalMetadataColumn(self.a) with self.assertRaisesRegex(ValueError, r'vmin must be less than.*\s\s' r'0\.5.*greater.*0\.0'): confusion_matrix(self.tmpd, self.a, b, vmin=.5, vmax=None) def test_confusion_matrix_vmax_too_low(self): b = qiime2.CategoricalMetadataColumn(self.a) with self.assertRaisesRegex(ValueError, r'vmax must be greater than.*' r'\s\s0\.5.*less.*1\.0'): confusion_matrix(self.tmpd, self.a, b, vmin=None, vmax=.5) def test_confusion_matrix_vmin_too_high_and_vmax_too_low(self): b = qiime2.CategoricalMetadataColumn(self.a) with self.assertRaisesRegex(ValueError, r'vmin must be less than.*\s' r'\s0\.5.*greater.*0\.0\s.*vmax must be ' r'greater than.*\s\s0\.5.*less.*1\.0'): confusion_matrix(self.tmpd, self.a, b, vmin=.5, vmax=.5) def test_confusion_matrix_dtype_coercion(self): predictions = pd.Series([1, 1, 1, 2, 2, 2], index=pd.Index(['a', 'b', 'c', 'd', 'e', 'f'], name='sample_id'), name='features') # NOTE: the targets are numbers but represented as str truth = qiime2.CategoricalMetadataColumn(pd.Series( ['1', '2', '1', '2', '1', '2'], index=pd.Index(['a', 'b', 'c', 'd', 'e', 'f'], name='sample-id'), name='target')) confusion_matrix(self.tmpd, predictions, truth) self.assertTrue('index.html' in listdir(self.tmpd)) # test confusion matrix plotting independently to see how it handles # partially overlapping classes when true labels are superset def test_predict_and_plot_true_labels_are_superset(self): b = pd.Series(['a', 'a', 'b', 'b', 'b', 'b'], name='site', index=['a1', 'a2', 'b1', 'b2', 'c1', 'c2']) exp = pd.DataFrame( [[1., 0., 0., ''], [0., 1., 0., ''], [0., 1., 0., ''], ['', '', '', 0.666666666], ['', '', '', 0.3333333333], ['', '', '', 2.]], columns=['a', 'b', 'c', 'Overall Accuracy'], index=['a', 'b', 'c', 'Overall Accuracy', 'Baseline Accuracy', 'Accuracy Ratio']) predictions, confusion = _predict_and_plot(self.tmpd, self.a, b) pdt.assert_frame_equal(exp, predictions) # test confusion matrix plotting independently to see how it handles # partially overlapping classes when true labels are superset def test_predict_and_plot_true_labels_are_subset(self): b = pd.Series(['a', 'a', 'b', 'b', 'c', 'd'], name='site', index=['a1', 'a2', 'b1', 'b2', 'c1', 'c2']) exp = pd.DataFrame( [[1., 0., 0., 0., ''], [0., 1., 0., 0., ''], [0., 0., 0.5, 0.5, ''], [0., 0., 0., 0., ''], ['', '', '', '', 0.8333333333], ['', '', '', '', 0.3333333333], ['', '', '', '', 2.5]], columns=['a', 'b', 'c', 'd', 'Overall Accuracy'], index=['a', 'b', 'c', 'd', 'Overall Accuracy', 'Baseline Accuracy', 'Accuracy Ratio']) predictions, confusion = _predict_and_plot(self.tmpd, self.a, b) pdt.assert_frame_equal(exp, predictions) # test confusion matrix plotting independently to see how it handles # partially overlapping classes when true labels are mutually exclusive def test_predict_and_plot_true_labels_are_mutually_exclusive(self): b = pd.Series(['a', 'a', 'e', 'e', 'd', 'd'], name='site', index=['a1', 'a2', 'b1', 'b2', 'c1', 'c2']) exp = pd.DataFrame( [[1., 0., 0., 0., 0., ''], [0., 0., 0., 0., 1., ''], [0., 0., 0., 1., 0., ''], [0., 0., 0., 0., 0., ''], [0., 0., 0., 0., 0., ''], ['', '', '', '', '', 0.3333333333], ['', '', '', '', '', 0.3333333333], ['', '', '', '', '', 1.]], columns=['a', 'b', 'c', 'd', 'e', 'Overall Accuracy'], index=['a', 'b', 'c', 'd', 'e', 'Overall Accuracy', 'Baseline Accuracy', 'Accuracy Ratio']) predictions, confusion = _predict_and_plot(self.tmpd, self.a, b) pdt.assert_frame_equal(exp, predictions) def test_scatterplot(self): b = qiime2.NumericMetadataColumn(self.c) scatterplot(self.tmpd, self.c, b) def test_add_sample_size_to_xtick_labels(self): labels = _add_sample_size_to_xtick_labels(self.a, ['a', 'b', 'c']) exp = ['a (n=2)', 'b (n=2)', 'c (n=2)'] self.assertListEqual(labels, exp) # now test performance when extra classes are present def test_add_sample_size_to_xtick_labels_extra_classes(self): labels = _add_sample_size_to_xtick_labels( self.a, [0, 'a', 'b', 'bb', 'c']) exp = ['0 (n=0)', 'a (n=2)', 'b (n=2)', 'bb (n=0)', 'c (n=2)'] self.assertListEqual(labels, exp) def test_match_series_or_die(self): exp = pd.Series(['a', 'b', 'c'], name='site', index=['a1', 'b2', 'c2']) exp.index.name = 'SampleID' a, b = _match_series_or_die(self.a, self.bogus, 'ignore') pdt.assert_series_equal(exp, a) pdt.assert_series_equal(exp, b) def test_match_series_or_die_missing_samples(self): with self.assertRaisesRegex(ValueError, "Missing samples"): a, b = _match_series_or_die(self.a, self.bogus, 'error')