def test_compute_splif_features_in_range(self):
prot_xyz, prot_rdk = rgf.load_molecule(self.protein_file)
lig_xyz, lig_rdk = rgf.load_molecule(self.ligand_file)
prot_num_atoms = prot_rdk.GetNumAtoms()
lig_num_atoms = lig_rdk.GetNumAtoms()
distance = rgf.compute_pairwise_distances(
protein_xyz=prot_xyz, ligand_xyz=lig_xyz)
for bins in ((0, 2), (2, 3)):
splif_dict = rgf.compute_splif_features_in_range(
prot_rdk,
lig_rdk,
distance,
bins,
)
self.assertIsInstance(splif_dict, dict)
for (prot_idx, lig_idx), ecfp_pair in splif_dict.items():
for idx in (prot_idx, lig_idx):
self.assertIsInstance(idx, (int, np.int64))
self.assertGreaterEqual(prot_idx, 0)
self.assertLess(prot_idx, prot_num_atoms)
self.assertGreaterEqual(lig_idx, 0)
self.assertLess(lig_idx, lig_num_atoms)
for ecfp in ecfp_pair:
ecfp_idx, ecfp_frag = ecfp.split(',')
ecfp_idx = int(ecfp_idx)
self.assertGreaterEqual(ecfp_idx, 0)
def test_compute_pairwise_distances(self):
n1 = 10
n2 = 50
coords1 = np.random.rand(n1, 3)
coords2 = np.random.rand(n2, 3)
distance = rgf.compute_pairwise_distances(coords1, coords2)
self.assertEqual(distance.shape, (n1, n2))
self.assertTrue((distance >= 0).all())
# random coords between 0 and 1, so the max possible distance in sqrt(2)
self.assertTrue((distance <= 2.0**0.5).all())
# check if correct distance metric was used
coords1 = np.array([[0, 0, 0], [1, 0, 0]])
coords2 = np.array([[1, 0, 0], [2, 0, 0], [3, 0, 0]])
distance = rgf.compute_pairwise_distances(coords1, coords2)
self.assertTrue((distance == [[1, 2, 3], [0, 1, 2]]).all())
def test_featurize_binding_pocket_ecfp(self):
prot_xyz, prot_rdk = rgf.load_molecule(self.protein_file)
lig_xyz, lig_rdk = rgf.load_molecule(self.ligand_file)
distance = rgf.compute_pairwise_distances(
protein_xyz=prot_xyz, ligand_xyz=lig_xyz)
# check if results are the same if we provide precomputed distances
prot_dict, lig_dict = rgf.featurize_binding_pocket_ecfp(
prot_xyz,
prot_rdk,
lig_xyz,
lig_rdk,
)
prot_dict_dist, lig_dict_dist = rgf.featurize_binding_pocket_ecfp(
prot_xyz, prot_rdk, lig_xyz, lig_rdk, pairwise_distances=distance)
# ...but first check if we actually got two dicts
self.assertIsInstance(prot_dict, dict)
self.assertIsInstance(lig_dict, dict)
self.assertEqual(prot_dict, prot_dict_dist)
self.assertEqual(lig_dict, lig_dict_dist)
# check if we get less features with smaller distance cutoff
prot_dict_d2, lig_dict_d2 = rgf.featurize_binding_pocket_ecfp(
prot_xyz,
prot_rdk,
lig_xyz,
lig_rdk,
cutoff=2.0,
)
prot_dict_d6, lig_dict_d6 = rgf.featurize_binding_pocket_ecfp(
prot_xyz,
prot_rdk,
lig_xyz,
lig_rdk,
cutoff=6.0,
)
self.assertLess(len(prot_dict_d2), len(prot_dict))
# ligands are typically small so all atoms might be present
self.assertLessEqual(len(lig_dict_d2), len(lig_dict))
self.assertGreater(len(prot_dict_d6), len(prot_dict))
self.assertGreaterEqual(len(lig_dict_d6), len(lig_dict))
# check if using different ecfp_degree changes anything
prot_dict_e3, lig_dict_e3 = rgf.featurize_binding_pocket_ecfp(
prot_xyz,
prot_rdk,
lig_xyz,
lig_rdk,
ecfp_degree=3,
)
self.assertNotEqual(prot_dict_e3, prot_dict)
self.assertNotEqual(lig_dict_e3, lig_dict)
def test_featurize_splif(self):
prot_xyz, prot_rdk = rgf.load_molecule(self.protein_file)
lig_xyz, lig_rdk = rgf.load_molecule(self.ligand_file)
distance = rgf.compute_pairwise_distances(
protein_xyz=prot_xyz, ligand_xyz=lig_xyz)
bins = [(1, 2), (2, 3)]
dicts = rgf.featurize_splif(
prot_xyz,
prot_rdk,
lig_xyz,
lig_rdk,
contact_bins=bins,
pairwise_distances=distance,
ecfp_degree=2)
expected_dicts = [
rgf.compute_splif_features_in_range(
prot_rdk, lig_rdk, distance, c_bin, ecfp_degree=2) for c_bin in bins
]
self.assertIsInstance(dicts, list)
self.assertEqual(dicts, expected_dicts)