def test_coulomb_matrix(self): """ Test CoulombMatrix. """ f = CoulombMatrix(self.num_atoms) rval = f([self.mol_with_no_conf]) assert rval.shape == (1, self.num_atoms, self.num_atoms) rval = f([self.mol_with_one_conf]) assert rval.shape == (1, self.num_atoms, self.num_atoms) rval = f([self.mol_with_multi_conf]) assert rval.shape == (1, self.num_confs, self.num_atoms, self.num_atoms)
def test_upper_tri_coulomb_matrix(self): """ Test upper triangular CoulombMatrix. """ f = CoulombMatrix(self.num_atoms, upper_tri=True) size = np.triu_indices(self.num_atoms)[0].size rval = f([self.mol_with_no_conf]) assert rval.shape == (1, size) rval = f([self.mol_with_one_conf]) assert rval.shape == (1, size) rval = f([self.mol_with_multi_conf]) assert rval.shape == (1, self.num_confs, size)
def test_coulomb_matrix_padding(self): """ Test CoulombMatrix with padding. """ max_atoms = self.num_atoms * 2 f = CoulombMatrix(max_atoms=max_atoms) rval = f([self.mol_with_no_conf]) assert rval.shape == (1, max_atoms, max_atoms) rval = f([self.mol_with_one_conf]) assert rval.shape == (1, max_atoms, max_atoms) rval = f([self.mol_with_multi_conf]) assert rval.shape == (1, self.num_confs, max_atoms, max_atoms)
def test_coulomb_matrix_hydrogens(self):
"""
Test no hydrogen removal.
"""
f = CoulombMatrix(
max_atoms=self.num_atoms, remove_hydrogens=False, upper_tri=True)
size = np.triu_indices(self.num_atoms)[0].size
rval = f([self.mol_with_no_conf])
assert rval.shape == (1, size)
rval = f([self.mol_with_one_conf])
assert rval.shape == (1, size)
rval = f([self.mol_with_multi_conf])
assert rval.shape == (1, self.num_confs, size)
def test_coulomb_matrix_no_hydrogens(self):
"""
Test hydrogen removal.
"""
num_atoms_with_no_H = self.mol_with_no_conf.GetNumAtoms()
assert num_atoms_with_no_H < self.num_atoms
f = CoulombMatrix(
max_atoms=num_atoms_with_no_H, remove_hydrogens=True, upper_tri=True)
size = np.triu_indices(num_atoms_with_no_H)[0].size
rval = f([self.mol_with_no_conf])
assert rval.shape == (1, size)
rval = f([self.mol_with_one_conf])
assert rval.shape == (1, size)
rval = f([self.mol_with_multi_conf])
assert rval.shape == (1, self.num_confs, size)
def generate_coulomb_matrices(self):
if not __deepchem_imported__:
self.log.error(
"deepchem was not imported or available in this execution environment"
)
sys.exit()
self.log.info(
"Generating Coulomb Matrices, may take a few moments ...")
matrix_featurizer = CoulombMatrix(self.max_atoms,
randomize=False,
n_samples=1)
eigen_featurizer = CoulombMatrixEig(self.max_atoms)
matrices = []
smiles = []
arrays = []
eigenarrays = []
num_confs = 1
for index, row in self.data.iterrows():
mol = Chem.MolFromSmiles(row["SMILE"])
Chem.AddHs(mol)
ids = AllChem.EmbedMultipleConfs(mol, numConfs=num_confs)
if len(ids) != num_confs:
ids = AllChem.EmbedMultipleConfs(mol,
numConfs=num_confs,
ignoreSmoothingFailures=True)
if len(ids) != num_confs:
self.log.warning(
"Coulomb Matrix - unable to generate %d conformer(s) for smile: %s",
num_confs, row["SMILE"])
if len(ids) == num_confs:
for id in ids:
AllChem.UFFOptimizeMolecule(mol, confId=id)
matrix = matrix_featurizer.coulomb_matrix(mol)
matrices.append(matrix)
arrays.append(matrix[0].flatten())
smiles.append(row["SMILE"])
eigenvalues = eigen_featurizer.featurize([mol])
eigenarrays.append(eigenvalues[0].flatten())
pd_dict = {
"SMILE": smiles,
"COULOMB": matrices,
"COULOMB_ARRAY": arrays,
"COULOMB_EIGEN": eigenarrays
}
coulomb_frame = pd.DataFrame(pd_dict)
before_ids = list(self.data["ID"])
self.data = pd.merge(self.data,
coulomb_frame,
how="inner",
on=["SMILE"])
self.data = self.data.drop_duplicates(subset=["SMILE"], keep="first")
after_ids = list(self.data["ID"])
missing_list = list(set(before_ids) - set(after_ids))
for missing in missing_list:
self.log.warning(
"Dropped molecule ID: %s after join with Coulomb Matrix data",
missing)