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_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)
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_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_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_compute_charge_dictionary(self):
for fname in (self.ligand_file, self.protein_file):
_, mol = rgf.load_molecule(fname)
ComputeGasteigerCharges(mol)
charge_dict = rgf.compute_charge_dictionary(mol)
self.assertEqual(len(charge_dict), mol.GetNumAtoms())
for i in range(mol.GetNumAtoms()):
self.assertIn(i, charge_dict)
self.assertIsInstance(charge_dict[i], (float, int))
def test_compute_charge_dictionary(self):
for fname in (self.ligand_file, self.protein_file):
_, mol = rgf.load_molecule(fname)
ComputeGasteigerCharges(mol)
charge_dict = rgf.compute_charge_dictionary(mol)
self.assertEqual(len(charge_dict), mol.GetNumAtoms())
for i in range(mol.GetNumAtoms()):
self.assertIn(i, charge_dict)
self.assertIsInstance(charge_dict[i], (float, int))
def test_voxelize(self):
prot_xyz, prot_rdk = rgf.load_molecule(self.protein_file)
lig_xyz, lig_rdk = rgf.load_molecule(self.ligand_file)
centroid = rgf.compute_centroid(lig_xyz)
prot_xyz = rgf.subtract_centroid(prot_xyz, centroid)
lig_xyz = rgf.subtract_centroid(lig_xyz, centroid)
prot_ecfp_dict, lig_ecfp_dict = rgf.featurize_binding_pocket_ecfp(
prot_xyz, prot_rdk, lig_xyz, lig_rdk)
box_w = 20
f_power = 5
rgf_featurizer = rgf.RdkitGridFeaturizer(
box_width=box_w,
ecfp_power=f_power,
feature_types=['all_combined'],
flatten=True,
sanitize=True)
prot_tensor = rgf_featurizer._voxelize(
rgf.convert_atom_to_voxel,
rgf.hash_ecfp,
prot_xyz,
feature_dict=prot_ecfp_dict,
channel_power=f_power)
self.assertEqual(prot_tensor.shape, tuple([box_w] * 3 + [2**f_power]))
all_features = prot_tensor.sum()
# protein is too big for the box, some features should be missing
self.assertGreater(all_features, 0)
self.assertLess(all_features, prot_rdk.GetNumAtoms())
lig_tensor = rgf_featurizer._voxelize(
rgf.convert_atom_to_voxel,
rgf.hash_ecfp,
lig_xyz,
feature_dict=lig_ecfp_dict,
channel_power=f_power)
self.assertEqual(lig_tensor.shape, tuple([box_w] * 3 + [2**f_power]))
all_features = lig_tensor.sum()
# whole ligand should fit in the box
self.assertEqual(all_features, lig_rdk.GetNumAtoms())
def test_load_molecule(self):
# adding hydrogens and charges is tested in dc.utils
for add_hydrogens in (True, False):
for calc_charges in (True, False):
mol_xyz, mol_rdk = rgf.load_molecule(self.ligand_file, add_hydrogens,
calc_charges)
num_atoms = mol_rdk.GetNumAtoms()
self.assertIsInstance(mol_xyz, np.ndarray)
self.assertIsInstance(mol_rdk, Mol)
self.assertEqual(mol_xyz.shape, (num_atoms, 3))
def setUp(self):
current_dir = os.path.dirname(os.path.realpath(__file__))
# simple flat ring
self.cycle4 = MolFromSmiles('C1CCC1')
self.cycle4.Compute2DCoords()
# load and sanitize two real molecules
_, self.prot = rgf.load_molecule(
os.path.join(current_dir, '3ws9_protein_fixer_rdkit.pdb'),
add_hydrogens=False,
calc_charges=False,
sanitize=True)
_, self.lig = rgf.load_molecule(
os.path.join(current_dir, '3ws9_ligand.sdf'),
add_hydrogens=False,
calc_charges=False,
sanitize=True)
def setUp(self):
current_dir = os.path.dirname(os.path.realpath(__file__))
# simple flat ring
self.cycle4 = MolFromSmiles('C1CCC1')
self.cycle4.Compute2DCoords()
# load and sanitize two real molecules
_, self.prot = rgf.load_molecule(os.path.join(
current_dir, '3ws9_protein_fixer_rdkit.pdb'),
add_hydrogens=False,
calc_charges=False,
sanitize=True)
_, self.lig = rgf.load_molecule(os.path.join(current_dir,
'3ws9_ligand.sdf'),
add_hydrogens=False,
calc_charges=False,
sanitize=True)
def test_voxelize(self):
prot_xyz, prot_rdk = rgf.load_molecule(self.protein_file)
lig_xyz, lig_rdk = rgf.load_molecule(self.ligand_file)
centroid = rgf.compute_centroid(lig_xyz)
prot_xyz = rgf.subtract_centroid(prot_xyz, centroid)
lig_xyz = rgf.subtract_centroid(lig_xyz, centroid)
prot_ecfp_dict, lig_ecfp_dict = rgf.featurize_binding_pocket_ecfp(
prot_xyz, prot_rdk, lig_xyz, lig_rdk)
box_w = 20
f_power = 5
rgf_featurizer = rgf.RdkitGridFeaturizer(
box_width=box_w,
ecfp_power=f_power,
feature_types=['all_combined'],
flatten=True,
sanitize=True)
prot_tensor = rgf_featurizer._voxelize(
rgf.convert_atom_to_voxel,
rgf.hash_ecfp,
prot_xyz,
feature_dict=prot_ecfp_dict,
channel_power=f_power)
self.assertEqual(prot_tensor.shape, tuple([box_w] * 3 + [2**f_power]))
all_features = prot_tensor.sum()
# protein is too big for the box, some features should be missing
self.assertGreater(all_features, 0)
self.assertLess(all_features, prot_rdk.GetNumAtoms())
lig_tensor = rgf_featurizer._voxelize(
rgf.convert_atom_to_voxel,
rgf.hash_ecfp,
lig_xyz,
feature_dict=lig_ecfp_dict,
channel_power=f_power)
self.assertEqual(lig_tensor.shape, tuple([box_w] * 3 + [2**f_power]))
all_features = lig_tensor.sum()
# whole ligand should fit in the box
self.assertEqual(all_features, lig_rdk.GetNumAtoms())
def test_load_molecule(self):
# adding hydrogens and charges is tested in dc.utils
for add_hydrogens in (True, False):
for calc_charges in (True, False):
mol_xyz, mol_rdk = rgf.load_molecule(self.ligand_file, add_hydrogens,
calc_charges)
num_atoms = mol_rdk.GetNumAtoms()
self.assertIsInstance(mol_xyz, np.ndarray)
self.assertIsInstance(mol_rdk, Mol)
self.assertEqual(mol_xyz.shape, (num_atoms, 3))
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)
def test_compute_all_ecfp(self):
_, mol = rgf.load_molecule(self.ligand_file)
num_atoms = mol.GetNumAtoms()
for degree in range(1, 4):
# TODO test if dict contains smiles
ecfp_all = rgf.compute_all_ecfp(mol, degree=degree)
self.assertIsInstance(ecfp_all, dict)
self.assertEqual(len(ecfp_all), num_atoms)
self.assertEqual(list(ecfp_all.keys()), list(range(num_atoms)))
num_ind = np.random.choice(range(1, num_atoms))
indices = list(np.random.choice(num_atoms, num_ind, replace=False))
ecfp_selected = rgf.compute_all_ecfp(mol, indices=indices, degree=degree)
self.assertIsInstance(ecfp_selected, dict)
self.assertEqual(len(ecfp_selected), num_ind)
self.assertEqual(sorted(ecfp_selected.keys()), sorted(indices))
def test_compute_all_ecfp(self):
_, mol = rgf.load_molecule(self.ligand_file)
num_atoms = mol.GetNumAtoms()
for degree in range(1, 4):
# TODO test if dict contains smiles
ecfp_all = rgf.compute_all_ecfp(mol, degree=degree)
self.assertIsInstance(ecfp_all, dict)
self.assertEqual(len(ecfp_all), num_atoms)
self.assertEqual(list(ecfp_all.keys()), list(range(num_atoms)))
num_ind = np.random.choice(range(1, num_atoms))
indices = list(np.random.choice(num_atoms, num_ind, replace=False))
ecfp_selected = rgf.compute_all_ecfp(mol, indices=indices, degree=degree)
self.assertIsInstance(ecfp_selected, dict)
self.assertEqual(len(ecfp_selected), num_ind)
self.assertEqual(sorted(ecfp_selected.keys()), sorted(indices))
