def test_sublist_from_qm9(self):
features_list, connection_list, global_list, \
index1_list, index2_list, targets = load_qm9_faber(
self.db_connection)
ids = ['000002', '000001']
t, new_features = sublist_from_qm9(ids, targets, features_list, connection_list)
self.assertEqual(t.shape[0], 2)
self.assertListEqual(list(t.index), ids)
def test_feature_clean(self):
features_list, connection_list, global_list, index1_list, index2_list, targets = load_qm9_faber(
self.db_connection)
fc = FeatureClean()
out = fc.fit_transform(features_list)
# first molecule has 5 atoms and final feature number is 21
self.assertListEqual(list(out[0].shape), [5, 21])
fc = FeatureClean(categorical=["bond_type", "same_ring"],
feature_labels=["bond_type", 'graph_distance', 'same_ring', "spatial_distance"])
out = fc.fit_transform(connection_list)
# note that the bond feature dimension is 27 for the whole dataset
# here we are missing 1 element (totally 5), thus the dimension is
# 4 + 1 + 1 + 20 = 26
self.assertListEqual(list(out[0].shape), [20, 26])
def test_load_qm9(self):
features_list, connection_list, global_list, index1_list, index2_list, targets = load_qm9_faber(
self.db_connection)
data = [features_list, connection_list, global_list, index1_list, index2_list, targets]
# 11 molecules
self.assertListEqual([len(i) for i in data[:-1]], [11] * 5)
# first molecule has 5 atoms and 20 bonds
self.assertEqual(len(features_list[0]), 5)
self.assertEqual(len(connection_list[0]), 20)
self.assertEqual(len(global_list[0]), 1)
self.assertEqual(len(index1_list[0]), 20)
self.assertEqual(len(index2_list[0]), 20)
# target dimension is 11*19
self.assertListEqual(list(targets.shape), [11, 19])
