def from_smiles(self, smiles, hydrogens_are_explicit=True):
"""
It initializes an RDKit's Molecule object from a SMILES tag.
Parameters
----------
smiles : str
The SMILES tag to construct the molecule structure with
hydrogens_are_explicit : bool
Whether the SMILES tag has explicit information about
hydrogen atoms or not. Otherwise, they will be added when
the molecule is built. Default is True
Returns
-------
molecule : an rdkit.Chem.rdchem.Mol object
The RDKit's Molecule object
"""
from rdkit.Chem import AllChem as Chem
molecule = Chem.MolFromSmiles(smiles, sanitize=False)
Chem.SanitizeMol(
molecule, Chem.SANITIZE_ALL ^ Chem.SANITIZE_ADJUSTHS
^ Chem.SANITIZE_SETAROMATICITY)
# Add hydrogens to molecule
if not hydrogens_are_explicit:
molecule = Chem.AddHs(molecule)
# Generate 3D coordinates
Chem.EmbedMolecule(molecule)
return molecule
def generate_structures(vae, smi, char_to_index, limit=1e4, write=False):
rdkit_mols = []
temps = []
iterations = []
iteration = limit_counter = 0
while True:
iteration += 1
limit_counter += 1
t = random.random()*2
candidate = decode_smiles(vae, smi, char_to_index, temp=t).split(" ")[0]
try:
sampled = Chem.MolFromSmiles(candidate)
cation = Chem.AddHs(sampled)
Chem.EmbedMolecule(cation, Chem.ETKDG())
Chem.UFFOptimizeMolecule(cation)
cation = Chem.RemoveHs(cation)
candidate = Chem.MolToSmiles(cation)
if candidate not in rdkit_mols:
temps.append(t)
iterations.append(iteration)
rdkit_mols.append(candidate)
limit_counter = 0
df = pd.DataFrame([rdkit_mols,temps,iterations]).T
df.columns = ['smiles', 'temperature', 'iteration']
print(df)
except:
pass
if limit_counter > limit:
break
if write:
df = pd.DataFrame([rdkit_mols,temps,iterations]).T
df.columns = ['smiles', 'temperature', 'iteration']
pd.DataFrame.to_csv(df, path_or_buf='{}.csv'.format(write), index=False)
return df
def from_smiles(self, smiles):
"""
It initializes an RDKit's Molecule object from a SMILES tag.
Parameters
----------
smiles : str
The SMILES tag to construct the molecule structure with.
Returns
-------
molecule : an rdkit.Chem.rdchem.Mol object
The RDKit's Molecule object
"""
from rdkit.Chem import AllChem as Chem
molecule = Chem.MolFromSmiles(smiles)
# Add hydrogens to molecule
molecule = Chem.AddHs(molecule)
# Generate 3D coordinates
Chem.EmbedMolecule(molecule)
return molecule
def _getMolWithEHTcharges(self, mol : Chem.Mol):
"""
Prepare the molecule and calculate the charges.
TODO: perhaps this should be divided into more methods for more flexibility...
:param mol:
:return:
"""
mol_ = Chem.RemoveHs(mol)
mol_ = Chem.AddHs(mol_)
Chem.EmbedMolecule(mol_)
# should create two output files ('run.out' and 'nul') TODO: maybe this is not really needed or even desirable in the final version?
passed, res = rdEHTTools.RunMol(mol_)
nat = len(mol_.GetAtoms())
charges = res.GetAtomicCharges()
if not self.includeHs:
mol_ = Chem.RemoveHs(mol)
# creates a mol object with charges set as double properties on atoms of that mol object
for i in range(nat):
if i < len(mol_.GetAtoms()):
mol_.GetAtomWithIdx(i).SetDoubleProp('EHTcharge', charges[i])
else:
break
return mol_
def get_descriptors(smiles):
"""
Get a dictionary of RDKit descriptors from a SMILES string.
Parameters
----------
smiles : str
The SMILES string of the chemical of interest
Returns
-------
descriptors : dict
A collection of molecular descriptors
Notes: Developed with RDKit 2019.03.4, although doc pages listed 2019.03.1
"""
mol = Chem.MolFromSmiles(smiles)
mol = Chem.AddHs(mol)
Chem.EmbedMolecule(mol, Chem.ETKDG())
descriptors = {}
# Starting with simple descriptors:
# https://www.rdkit.org/docs/source/rdkit.Chem.Descriptors.html
# Molecular weight
descriptors['molwt'] = Descriptors.ExactMolWt(mol)
# Partial charge metrics
descriptors['max_abs_partial_charge'] = Descriptors.MaxAbsPartialCharge(mol)
descriptors['max_partial_charge'] = Descriptors.MaxPartialCharge(mol)
descriptors['min_abs_partial_charge'] = Descriptors.MinAbsPartialCharge(mol)
descriptors['min_partial_charge'] = Descriptors.MinPartialCharge(mol)
# Basic electron counts
descriptors['num_radical_electrons'] = Descriptors.NumRadicalElectrons(mol)
descriptors['num_valence_electrons'] = Descriptors.NumValenceElectrons(mol)
# 3-D descriptors
# https://www.rdkit.org/docs/source/rdkit.Chem.Descriptors3D.html
# Calculating these should produce the same result, according to some basic tests
# descriptors['asphericity'] = rdMolDescriptors.CalcAsphericity(mol)
# descriptors['eccentricity'] = rdMolDescriptors.CalcEccentricity(mol)
descriptors['asphericity'] = Descriptors3D.Asphericity(mol)
descriptors['eccentricity'] = Descriptors3D.Eccentricity(mol)
descriptors['inertial_shape_factor'] = Descriptors3D.InertialShapeFactor(mol)
descriptors['radius_of_gyration'] = Descriptors3D.RadiusOfGyration(mol)
descriptors['spherocity_index'] = Descriptors3D.SpherocityIndex(mol)
# Graph descriptors
# https://www.rdkit.org/docs/source/rdkit.Chem.GraphDescriptors.html
descriptors['balaban_j'] = GraphDescriptors.BalabanJ(mol)
descriptors['bertz_ct'] = GraphDescriptors.BertzCT(mol)
descriptors['chi0'] = GraphDescriptors.Chi0(mol)
descriptors['chi0n'] = GraphDescriptors.Chi0n(mol)
descriptors['chi0v'] = GraphDescriptors.Chi0v(mol)
descriptors['chi1'] = GraphDescriptors.Chi1(mol)
descriptors['chi1n'] = GraphDescriptors.Chi1n(mol)
descriptors['chi1v'] = GraphDescriptors.Chi1v(mol)
descriptors['chi2n'] = GraphDescriptors.Chi2n(mol)
descriptors['chi2v'] = GraphDescriptors.Chi2v(mol)
descriptors['chi3n'] = GraphDescriptors.Chi3n(mol)
descriptors['chi3v'] = GraphDescriptors.Chi3v(mol)
descriptors['chi4n'] = GraphDescriptors.Chi4n(mol)
descriptors['chi4v'] = GraphDescriptors.Chi4v(mol)
descriptors['hall_kier_alpha'] = GraphDescriptors.HallKierAlpha(mol)
descriptors['kappa1'] = GraphDescriptors.Kappa1(mol)
descriptors['kappa2'] = GraphDescriptors.Kappa2(mol)
descriptors['kappa3'] = GraphDescriptors.Kappa3(mol)
# Predicted properties from Wildman and Crippen
descriptors['log_p'] = Descriptors.MolLogP(mol)
descriptors['refractivity'] = Descriptors.MolMR(mol)
return descriptors
