def testIncludeOnly(self):
cfac = ChemicalFeatures.BuildFeatureFactory(
os.path.join(RDConfig.RDBaseDir, 'Code', 'GraphMol', 'MolChemicalFeatures', 'test_data',
'featDef.txt'))
self.failUnless(cfac.GetNumFeatureDefs() == 2)
mol = Chem.MolFromSmiles("COCN")
rdDistGeom.EmbedMolecule(mol)
self.failUnless(cfac.GetNumMolFeatures(mol, includeOnly="HBondAcceptor") == 2)
self.failUnless(cfac.GetNumMolFeatures(mol, includeOnly="HBondDonor") == 1)
self.failUnless(cfac.GetNumMolFeatures(mol, includeOnly="Bogus") == 0)
self.failUnlessRaises(IndexError, lambda: cfac.GetMolFeature(mol, 1, includeOnly="HBondDonor"))
self.failUnlessRaises(IndexError,
lambda: cfac.GetMolFeature(mol, 2, includeOnly="HBondAcceptor"))
f = cfac.GetMolFeature(mol, 0, includeOnly="HBondDonor")
self.failUnless(f.GetFamily() == 'HBondDonor')
feats = cfac.GetFeaturesForMol(mol, includeOnly="HBondAcceptor")
self.failUnless(len(feats) == 2)
feats = cfac.GetFeaturesForMol(mol, includeOnly="HBondDonor")
self.failUnless(len(feats) == 1)
feats = cfac.GetFeaturesForMol(mol, includeOnly="Bogus")
self.failUnless(len(feats) == 0)
def setUp(self):
fdefFile = os.path.join(RDConfig.RDCodeDir, 'Chem', 'Pharm2D',
'test_data', 'BaseFeatures.fdef')
featFactory = ChemicalFeatures.BuildFeatureFactory(fdefFile)
self.factory = SigFactory.SigFactory(featFactory, minPointCount=2, maxPointCount=3)
self.factory.SetBins([(0, 2), (2, 5), (5, 8)])
self.factory.Init()
def _getFeatureFamily(mol):
FEATURE_DEF_FILE = os.path.join(RDConfig.RDDataDir, 'BaseFeatures.fdef')
feat_factory = ChemicalFeatures.BuildFeatureFactory(FEATURE_DEF_FILE)
hmol = rdkit.Chem.AddHs(mol)
AllChem.EmbedMolecule(hmol, useRandomCoords=True)
rc = rdkit.Chem.AllChem.EmbedMolecule(hmol)
logging.debug("Getting features for mol " + mol.GetProp("_Name"))
if rc < 0:
rc = rdkit.Chem.AllChem.EmbedMolecule(hmol, useRandomCoords=True)
if rc == 0:
try:
if rdkit.Chem.AllChem.UFFOptimizeMolecule(hmol) != 0:
rdkit.Chem.AllChem.UFFOptimizeMolecule(hmol, maxIters=1000)
except ValueError:
logging.error("Problem with 3D version of molecule " +
hmol.GetProp("_Name"))
pass
feats = feat_factory.GetFeaturesForMol(hmol)
atomFeatures = [["" for feature in range(len(feats))]
for atom in range(hmol.GetNumAtoms())]
for feature in feats:
for atomId in feature.GetAtomIds():
if feature.GetFamily() not in atomFeatures[atomId]:
atomFeatures[atomId].append(feature.GetFamily())
return atomFeatures
def processArgs(args, parser):
try:
factory = ChemicalFeatures.BuildFeatureFactory(args.fdefFilename)
except Exception:
parser.error(
"Could not parse Fdef file {0.fdefFilename}.".format(args))
with open(args.smilesFilename) as inF:
for lineNo, line in enumerate(inF, 1):
if lineNo == args.maxLines + 1:
break
smi = splitExpr.split(line.strip())[0].strip()
mol = Chem.MolFromSmiles(smi)
if mol is None:
logger.warning("Could not process smiles '%s' on line %d." %
(smi, lineNo))
continue
print('Mol-%d\t%s' % (lineNo, smi))
if args.reverseIt:
feats = factory.GetFeaturesForMol(mol)
for feat in feats:
print('\t%s-%s: ' % (feat.GetFamily(), feat.GetType()),
end='')
print(', '.join([str(x) for x in feat.GetAtomIds()]))
else:
featInfo = GetAtomFeatInfo(factory, mol)
for i, v in enumerate(featInfo):
print('\t% 2s(%d)' %
(mol.GetAtomWithIdx(i).GetSymbol(), i + 1),
end='')
if v:
print('\t', ', '.join(v))
else:
print()
def _align_molecules(self, molecules: List[Chem.Mol]) -> None:
""" Align a list of molecules to a given pharmacophore.
Parameters
----------
molecules : list of rdkit.Chem.Mol
List of molecules to align.
"""
self.n_molecules += len(molecules)
rdkit_pharmacophore, radii = self.pharmacophore.to_rdkit()
apply_radii_to_bounds(radii, rdkit_pharmacophore)
fdef = os.path.join(RDConfig.RDDataDir, 'BaseFeatures.fdef')
featFactory = ChemicalFeatures.BuildFeatureFactory(fdef)
MolScore = namedtuple("MolScore", ["score", "id", "mol"])
for mol in tqdm(molecules):
bounds_matrix = rdDistGeom.GetMoleculeBoundsMatrix(mol)
can_match, all_matches = EmbedLib.MatchPharmacophoreToMol(
mol, featFactory, rdkit_pharmacophore)
if can_match:
failed, _, matched_mols, _ = EmbedLib.MatchPharmacophore(
all_matches,
bounds_matrix,
rdkit_pharmacophore,
useDownsampling=True)
if failed:
matched_mol = MolScore(0.0, mol.GetProp("_Name"), mol)
self.molecules.append(matched_mol)
continue
else:
matched_mol = MolScore(0.0, mol.GetProp("_Name"), mol)
self.molecules.append(matched_mol)
continue
atom_match = [list(x.GetAtomIds()) for x in matched_mols]
try:
mol_H = Chem.AddHs(mol)
_, embeddings, _ = EmbedLib.EmbedPharmacophore(
mol_H, atom_match, rdkit_pharmacophore, count=10)
except:
continue
SSDs = transform_embeddings(rdkit_pharmacophore, embeddings,
atom_match)
if len(SSDs) == 0:
matched_mol = MolScore(0.0, mol.GetProp("_Name"), mol)
self.molecules.append(matched_mol)
continue
best_fit_index = min(enumerate(SSDs), key=itemgetter(1))[0]
score = 1 / SSDs[best_fit_index]
matched_mol = MolScore(score, mol.GetProp("_Name"),
embeddings[best_fit_index])
self.molecules.append(matched_mol)
def testGithub2603(self):
cfac = ChemicalFeatures.BuildFeatureFactory(
os.path.join(RDConfig.RDDataDir, "BaseFeatures.fdef"))
m = Chem.MolFromSmiles('OCc1ccccc1CN')
feats = cfac.GetFeaturesForMol(m)
self.assertEqual(feats[0].GetFamily(), 'Donor')
cfac = None
self.assertEqual(feats[0].GetFamily(), 'Donor')
def alchemy_nodes(mol):
"""Featurization for all atoms in a molecule. The atom indices
will be preserved.
Parameters
----------
mol : rdkit.Chem.rdchem.Mol
RDKit molecule object
Returns
-------
atom_feats_dict : dict
Dictionary for atom features
"""
atom_feats_dict = defaultdict(list)
is_donor = defaultdict(int)
is_acceptor = defaultdict(int)
fdef_name = osp.join(RDConfig.RDDataDir, 'BaseFeatures.fdef')
mol_featurizer = ChemicalFeatures.BuildFeatureFactory(fdef_name)
mol_feats = mol_featurizer.GetFeaturesForMol(mol)
mol_conformers = mol.GetConformers()
assert len(mol_conformers) == 1
for i in range(len(mol_feats)):
if mol_feats[i].GetFamily() == 'Donor':
node_list = mol_feats[i].GetAtomIds()
for u in node_list:
is_donor[u] = 1
elif mol_feats[i].GetFamily() == 'Acceptor':
node_list = mol_feats[i].GetAtomIds()
for u in node_list:
is_acceptor[u] = 1
num_atoms = mol.GetNumAtoms()
for u in range(num_atoms):
atom = mol.GetAtomWithIdx(u)
atom_type = atom.GetAtomicNum()
num_h = atom.GetTotalNumHs()
atom_feats_dict['node_type'].append(atom_type)
h_u = []
h_u += atom_type_one_hot(atom, ['H', 'C', 'N', 'O', 'F', 'S', 'Cl'])
h_u.append(atom_type)
h_u.append(is_acceptor[u])
h_u.append(is_donor[u])
h_u += atom_is_aromatic(atom)
h_u += atom_hybridization_one_hot(atom, [Chem.rdchem.HybridizationType.SP,
Chem.rdchem.HybridizationType.SP2,
Chem.rdchem.HybridizationType.SP3])
h_u.append(num_h)
atom_feats_dict['n_feat'].append(F.tensor(np.asarray(h_u, dtype=np.float32)))
atom_feats_dict['n_feat'] = F.stack(atom_feats_dict['n_feat'], dim=0)
atom_feats_dict['node_type'] = F.tensor(
np.asarray(atom_feats_dict['node_type'], dtype=np.int64))
return atom_feats_dict
def test4Search(self):
featFactory = ChemicalFeatures.BuildFeatureFactory(os.path.join(self.dataDir,
'BaseFeatures.fdef'))
activeFeats = [ChemicalFeatures.FreeChemicalFeature('Acceptor',
Geometry.Point3D(0.0, 0.0, 0.0)),
ChemicalFeatures.FreeChemicalFeature('Donor',
Geometry.Point3D(0.0, 0.0, 0.0)),
ChemicalFeatures.FreeChemicalFeature('Aromatic',
Geometry.Point3D(0.0, 0.0, 0.0))]
pcophore= Pharmacophore.Pharmacophore(activeFeats)
pcophore.setLowerBound(0,1,2.251)
pcophore.setUpperBound(0,1,2.451)
pcophore.setUpperBound2D(0,1,3)
pcophore.setLowerBound(0,2,4.970)
pcophore.setUpperBound(0,2,5.170)
pcophore.setUpperBound2D(0,2,6)
pcophore.setLowerBound(1,2,2.681)
pcophore.setUpperBound(1,2,2.881)
pcophore.setUpperBound2D(1,2,6)
inF = gzip.open(os.path.join(self.dataDir,'cdk2-syn-clip100.pkl.gz'),'rb')
nDone = 0
nMatches = 0
nHits = 0
while 1:
try:
name,molPkl,boundsMat = cPickle.load(inF, encoding='latin1')
if PY3:
molPkl = bytes(molPkl, encoding='latin1')
except:
break
nDone += 1
mol = Chem.Mol(molPkl)
boundsMat = rdDistGeom.GetMoleculeBoundsMatrix(mol)
DG.DoTriangleSmoothing(boundsMat)
canMatch,matches = EmbedLib.MatchPharmacophoreToMol(mol,featFactory,
pcophore)
if canMatch:
nMatches+=1
r = EmbedLib.MatchPharmacophore(matches,boundsMat,pcophore,
useDownsampling=True,use2DLimits=True,
mol=mol)
failed,bm,match,details = r
if not failed:
nHits+=1
self.assertEqual(nDone,100)
self.assertEqual(nMatches,93)
#print 'nhits:',nHits
self.assertEqual(nHits,67)
def __config_feature_factory(self):
"""
Initialize the 'feature factory' rdkit module with the
current molecule.
"""
fdefName = os.path.join(RDConfig.RDDataDir, 'BaseFeatures.fdef')
self.__feat_factory = ChemicalFeatures.BuildFeatureFactory(fdefName)
return
def str2molgraph(
rawstr, length
): # rawstr :tuple() e.g. ('<RX_6>', 'N', 'c', '1', 'n', 'c', '2', '[', 'n', 'H', ']', 'c', '(', 'C', 'C', 'C', 'c', '3', 'c', 's', 'c', '(', 'C', '(', '=', 'O', ')', 'O', ')', 'c', '3', ')', 'c', 'c', '2', 'c', '(', '=', 'O', ')', '[', 'n', 'H', ']', '1')
smiles = ''.join(rawstr[:length])
m = Chem.MolFromSmiles(smiles)
g = nx.Graph()
fdef_name = os.path.join(RDConfig.RDDataDir, 'BaseFeatures.fdef')
factory = ChemicalFeatures.BuildFeatureFactory(fdef_name)
feats = factory.GetFeaturesForMol(m)
atom_true_index = {}
atom_index = 0
# Nodes
for i in range(len(rawstr)):
if not need_emb(rawstr[i], EMB_ATOMS):
g.add_node(i)
else:
atom_true_index[
atom_index] = i # meanwhile, set a map dict to find the true index of atoms
atom_i = m.GetAtomWithIdx(atom_index)
atom_index += 1
g.add_node(i,
a_type=atom_i.GetSymbol(),
a_num=atom_i.GetAtomicNum(),
acceptor=0,
donor=0,
aromatic=atom_i.GetIsAromatic(),
hybridization=atom_i.GetHybridization(),
num_h=atom_i.GetTotalNumHs())
# Donor and Acceptor properties
for i in range(0, len(feats)):
if feats[i].GetFamily() == 'Donor':
node_list = feats[i].GetAtomIds()
for i in node_list:
if i in atom_true_index:
g.nodes[atom_true_index[i]]['donor'] = 1
elif feats[i].GetFamily() == 'Acceptor':
node_list = feats[i].GetAtomIds()
for i in node_list:
if i in atom_true_index:
g.nodes[atom_true_index[i]]['acceptor'] = 1
#Edges
for i in range(0, m.GetNumAtoms()):
for j in range(0, m.GetNumAtoms()):
e_ij = m.GetBondBetweenAtoms(i, j)
if e_ij is not None and i in atom_true_index and j in atom_true_index:
g.add_edge(atom_true_index[i],
atom_true_index[j],
b_type=e_ij.GetBondType())
return g
def get_factory(self):
"""
Generate the Ph4 feature factory
:return:
"""
if self.factory is None:
this_dir, this_filename = os.path.split(__file__)
data_path = os.path.join(this_dir, "data", "RDKitPh4.fdef")
self.factory = ChemicalFeatures.BuildFeatureFactory(data_path)
return self.factory
def test4Github252(self):
fdef = os.path.join(RDConfig.RDDataDir, 'BaseFeatures.fdef')
feat_factory = ChemicalFeatures.BuildFeatureFactory(fdef)
m1 = Chem.MolFromSmiles('Cc1ccccc1')
feats = feat_factory.GetFeaturesForMol(m1)
self.assertRaises(RuntimeError, lambda: Pharmacophore.Pharmacophore(feats))
AllChem.Compute2DCoords(m1)
Pharmacophore.Pharmacophore(feats)
def numpy_pp_fps(mols):
""" Calculate Gobbi and Poppinger pharmacophore fingerprints and return them as numpy.ndarrays
:param mols: {list} list of molecules (RDKit mols)
:return: numpy array containing row-wise fingerprints for every molecule
"""
feat_fact = ChemicalFeatures.BuildFeatureFactory()
sig_fact = SigFactory(feat_fact, useCounts=False, minPointCount=2, maxPointCount=3)
sig_fact.SetBins([(0, 2), (2, 4), (4, 6), (6, 8), (8, 100)])
sig_fact.Init()
return _rdk2numpy([Generate.Gen2DFingerprint(m, sig_fact) for m in mols if m])
def get_instance(cls):
try:
from rdkit import RDConfig
from rdkit.Chem import ChemicalFeatures
except ModuleNotFoundError:
raise ValueError("This class requires RDKit to be installed.")
if not cls._instance:
fdefName = os.path.join(RDConfig.RDDataDir, 'BaseFeatures.fdef')
cls._instance = ChemicalFeatures.BuildFeatureFactory(fdefName)
return cls._instance
def rdkit_featuredefinition() -> ChemicalFeatures.MolChemicalFeatureFactory:
""" Loads rdkit chemical feature factory.
Returns
-------
rdkit.Chem.rdMolChemicalFeatures.MolChemicalFeatureFactory
The feature factory.
"""
fdefName = os.path.join(RDConfig.RDDataDir, 'BaseFeatures.fdef')
return ChemicalFeatures.BuildFeatureFactory(fdefName)
def __init__(self, atms: typing.List[str]):
self.atms_to_idx = dict(zip(atms, range(len(atms))))
self.number_atom_options = len(self.atms_to_idx)
self.hyb_mapping = {Chem.rdchem.HybridizationType.SP:0 ,
Chem.rdchem.HybridizationType.SP2: 1,
Chem.rdchem.HybridizationType.SP3: 2}
self.number_hyb_options = len(self.hyb_mapping)
self.fdef_name = os.path.join(RDDataDir, 'BaseFeatures.fdef')
self.feats_factory = ChemicalFeatures.BuildFeatureFactory(self.fdef_name)
def __call__(self, mol):
"""Featurizes the input molecule.
Parameters
----------
mol : rdkit.Chem.rdchem.Mol
RDKit molecule instance.
Returns
-------
dict
Mapping atom_data_field as specified in the input argument to the atom
features, which is a float32 tensor of shape (N, M), N is the number of
atoms and M is the feature size.
"""
atom_features = []
AllChem.ComputeGasteigerCharges(mol)
num_atoms = mol.GetNumAtoms()
# Get information for donor and acceptor
fdef_name = osp.join(RDConfig.RDDataDir, 'BaseFeatures.fdef')
mol_featurizer = ChemicalFeatures.BuildFeatureFactory(fdef_name)
mol_feats = mol_featurizer.GetFeaturesForMol(mol)
is_donor, is_acceptor = self.get_donor_acceptor_info(mol_feats)
# Get a symmetrized smallest set of smallest rings
# Following the practice from Chainer Chemistry (https://github.com/chainer/
# chainer-chemistry/blob/da2507b38f903a8ee333e487d422ba6dcec49b05/chainer_chemistry/
# dataset/preprocessors/weavenet_preprocessor.py)
sssr = Chem.GetSymmSSSR(mol)
for i in range(num_atoms):
atom = mol.GetAtomWithIdx(i)
# Features that can be computed directly from RDKit atom instances, which is a list
feats = self._featurizer(atom)
# Donor/acceptor indicator
feats.append(float(is_donor[i]))
feats.append(float(is_acceptor[i]))
# Count the number of rings the atom belongs to for ring size between 3 and 8
count = [0 for _ in range(3, 9)]
for ring in sssr:
ring_size = len(ring)
if i in ring and 3 <= ring_size <= 8:
count[ring_size - 3] += 1
feats.extend(count)
atom_features.append(feats)
atom_features = np.stack(atom_features)
return {
self._atom_data_field:
F.zerocopy_from_numpy(atom_features.astype(np.float32))
}
def construct_hydrogen_bonding(mol, num_max_atoms=WEAVE_DEFAULT_NUM_MAX_ATOMS):
fdefName = os.path.join(RDConfig.RDDataDir, 'BaseFeatures.fdef')
factory = ChemicalFeatures.BuildFeatureFactory(fdefName)
feats = factory.GetFeaturesForMol(mol)
hydrogen_bonding_vec = numpy.zeros((num_max_atoms, 2), dtype=numpy.float32)
for f in feats:
if f.GetFamily() == 'Donor':
idx = f.GetAtomIds()[0]
hydrogen_bonding_vec[idx, 0] = 1.0
if f.GetFamily() == 'Acceptor':
idx = f.GetAtomIds()[0]
hydrogen_bonding_vec[idx, 1] = 1.0
return hydrogen_bonding_vec
def load_factory(filename=None):
"""
Loads RDKit factory with custom feature patterns from a file.
:param filename: file name of fdef format file. If None the default patterns will be loaded. Default: None.
:type filename: str
:return: object of MolChemicalFeatureFactory class
"""
if filename is None:
filename = path.join(path.abspath(path.dirname(__file__)),
'smarts_features.fdef')
return ChemicalFeatures.BuildFeatureFactory(filename)
def __new__(cls):
if cls._instance is None:
cls._instance = super(
_ChemicalFeatureGenerator, cls).__new__(cls)
from rdkit import RDConfig
from rdkit.Chem import ChemicalFeatures
fdef_path = path.join(RDConfig.RDDataDir, 'BaseFeatures.fdef')
cls._instance.feature_factory = \
ChemicalFeatures.BuildFeatureFactory(fdef_path)
return cls._instance
def extract_features(mol):
factory = ChemicalFeatures.BuildFeatureFactory('./LigityFeatures.fdef')
feats = factory.GetFeaturesForMol(mol)
features = []
for feat in feats:
feature = feat.GetFamily()
if feature in pharmacophores:
id = feat.GetId()
x, y, z = list(feat.GetPos())
string = str(id) + ',' + feature + ',' + str(x) + ',' + str(
y) + ',' + str(z)
features.append(string)
return features
def BuildSigFactory(options=None, fdefFile=None,
bins=[(2, 3), (3, 4), (4, 5), (5, 6), (6, 7), (7, 8), (8, 100)],
skipFeats=('LumpedHydrophobe', 'ZnBinder')):
if options:
fdefFile = options.fdefFile
if not fdefFile:
raise ValueError('bad fdef file')
from rdkit.Chem import ChemicalFeatures
from rdkit.Chem.Pharm2D import SigFactory
featFactory = ChemicalFeatures.BuildFeatureFactory(fdefFile)
sigFactory = SigFactory.SigFactory(featFactory, skipFeats=skipFeats, trianglePruneBins=False)
sigFactory.SetBins(bins)
return sigFactory
def get_node_features(self, mol):
fdef_name = osp.join(RDConfig.RDDataDir, 'BaseFeatures.fdef')
factory = ChemicalFeatures.BuildFeatureFactory(fdef_name)
x = []
sp = []
sp2 = []
sp3 = []
donor = []
acceptor = []
type_idx = []
for i in range(4):
x.append([])
for atom in mol.GetAtoms():
donor.append(0)
acceptor.append(0)
#print(atom.GetHybridization())
x[0].append(atom.GetAtomicNum())
x[1].append(atom.GetTotalValence())
x[2].append(atom.GetIsAromatic())
x[3].append(atom.GetTotalNumHs(includeNeighbors=True))
hybridization = atom.GetHybridization()
sp.append(1 if hybridization == HybridizationType.SP else 0)
sp2.append(1 if hybridization == HybridizationType.SP2 else 0)
sp3.append(1 if hybridization == HybridizationType.SP3 else 0)
type_idx.append(atom.GetAtomicNum())
# Now calculate donors and acceptors
feats = factory.GetFeaturesForMol(mol)
for j in range(0, len(feats)):
if feats[j].GetFamily() == 'Donor':
node_list = feats[j].GetAtomIds()
for k in node_list:
donor[k] = 1
elif feats[j].GetFamily() == 'Acceptor':
node_list = feats[j].GetAtomIds()
for k in node_list:
acceptor[k] = 1
# Now get the elements and make one hot encoding
x_element = []
for i in range(len(self.atomicNbrs)):
x_element.append([])
for i in range(len(type_idx)):
for j in range(len(self.atomicNbrs)):
if type_idx[i] == self.atomicNbrs[j]:
x_element[j].append(1)
else:
x_element[j].append(0)
# Now add them all
x = x + [donor] + [acceptor] + [sp] + [sp2] + [sp3] + x_element
x = np.array(x).T
#x = torch.tensor(x, dtype=torch.float)
return (x, x.sum(axis=0), len(x))
def smile_to_graph(smile):
mol = Chem.MolFromSmiles(smile)
mol = Chem.AddHs(mol)
fdef_name = os.path.join(RDConfig.RDDataDir, 'BaseFeatures.fdef')
factory = ChemicalFeatures.BuildFeatureFactory(fdef_name)
feats = factory.GetFeaturesForMol(mol)
#Construction of the graph
graph = Molecule()
graph.atoms = []
graph.bonds = []
for i in range(0, mol.GetNumAtoms()):
atom = mol.GetAtomWithIdx(i)
node = Atom(i,
atom.GetSymbol(),
atom.GetAtomicNum(),
aromatic=atom.GetIsAromatic(),
hybrid=atom.GetHybridization(),
nbH=atom.GetTotalNumHs())
graph.add_atom(node)
for i in range(0, len(feats)):
if feats[i].GetFamily() == 'Donor':
node_list = feats[i].GetAtomIds()
for i in node_list:
graph.atoms[i].don = 1
elif feats[i].GetFamily() == 'Acceptor':
node_list = feats[i].GetAtomIds()
for i in node_list:
graph.atoms[i].acc = 1
for i in range(0, mol.GetNumAtoms()):
for j in range(0, mol.GetNumAtoms()):
e_ij = mol.GetBondBetweenAtoms(i, j)
if e_ij is not None:
bond = e_ij.GetBondTypeAsDouble()
graph.add_bond(i, j, bond)
return graph
def alchemy_nodes(self, mol):
atom_feats_dict, is_donor, is_acceptor = defaultdict(
list), defaultdict(int), defaultdict(int)
ddir = rdf.RDDataDir
mol_featurizer = cf.BuildFeatureFactory(osp.join(ddir, BASE_FT))
mol_conformers, mol_feats = mol.GetConformers(
), mol_featurizer.GetFeaturesForMol(mol)
geom = mol_conformers[0].GetPositions()
for i in range(len(mol_feats)):
if mol_feats[i].GetFamily() == 'Acceptor':
node_list = mol_feats[i].GetAtomIds()
for u in node_list:
is_acceptor[u] = 1
elif mol_feats[i].GetFamily() == 'None':
continue
elif mol_feats[i].GetFamily() == 'Donor':
node_list = mol_feats[i].GetAtomIds()
for u in node_list:
is_donor[u] = 1
num_atoms = mol.GetNumAtoms()
for u in range(num_atoms):
atom = mol.GetAtomWithIdx(u)
h_u = []
symbol, atom_type, aromatic = atom.GetSymbol(), atom.GetAtomicNum(
), atom.GetIsAromatic()
atom_feats_dict['node_type'].append(atom_type)
hybridization, num_h = atom.GetHybridization(), atom.GetTotalNumHs(
)
atom_feats_dict[POSITION].append(torch.FloatTensor(geom[u]))
h_u = h_u + [int(symbol == x) for x in ATOMS]
h_u.append(atom_type)
h_u.append(int(aromatic))
h_u.append(is_donor[u])
h_u.append(is_acceptor[u])
h_u += [
int(hybridization == x)
for x in (Chem.rdchem.HybridizationType.SP,
Chem.rdchem.HybridizationType.SP2,
Chem.rdchem.HybridizationType.SP3)
]
h_u.append(num_h)
atom_feats_dict[N_FEAT].append(torch.FloatTensor(h_u))
atom_feats_dict['node_type'] = torch.LongTensor(
atom_feats_dict['node_type'])
atom_feats_dict[N_FEAT] = torch.stack(atom_feats_dict[N_FEAT], dim=0)
atom_feats_dict[POSITION] = torch.stack(atom_feats_dict['pos'], dim=0)
return atom_feats_dict
def rawsmiles2graph(smiles):
# smiles = smiles.strip().replace(' ','')
m = Chem.MolFromSmiles(smiles)
g = nx.Graph()
fdef_name = os.path.join(RDConfig.RDDataDir, 'BaseFeatures.fdef')
factory = ChemicalFeatures.BuildFeatureFactory(fdef_name)
feats = factory.GetFeaturesForMol(m)
# Nodes
for i in range(0, m.GetNumAtoms()):
atom_i = m.GetAtomWithIdx(i)
g.add_node(i,
a_type=atom_i.GetSymbol(),
a_num=atom_i.GetAtomicNum(),
acceptor=0,
donor=0,
aromatic=atom_i.GetIsAromatic(),
hybridization=atom_i.GetHybridization(),
num_h=atom_i.GetTotalNumHs())
# Donor and Acceptor properties
for i in range(0, len(feats)):
if feats[i].GetFamily() == 'Donor':
node_list = feats[i].GetAtomIds()
for i in node_list:
g.node[i]['donor'] = 1
elif feats[i].GetFamily() == 'Acceptor':
node_list = feats[i].GetAtomIds()
for i in node_list:
g.node[i]['acceptor'] = 1
# Edges
for i in range(0, m.GetNumAtoms()):
for j in range(0, m.GetNumAtoms()):
e_ij = m.GetBondBetweenAtoms(i, j)
if e_ij is not None:
g.add_edge(i, j, b_type=e_ij.GetBondType())
else:
# Unbonded
g.add_edge(
i,
j,
b_type=None,
)
return g
def get_atom_features(mol, dist_matrix):
"""
Compute the following features for each atom in 'mol':
- atom type: H, C, N, O, F (one-hot)
- degree: 1, 2, 3, 4, 5 (one-hot)
- Hybridization: SP, SP2, SP3, UNSPECIFIED (one-hot)
- is aromatic: bool {0, 1}
- formal charge: int
- atomic number: float
- average bond length: float
- average weight of neigboring atoms: float
- donor: bool {0, 1}
- acceptor: bool {0, 1}
"""
n_atoms = mol.GetNumAtoms()
features = np.zeros((n_atoms, C.N_ATOM_FEATURES))
adj_matrix = rdmolops.GetAdjacencyMatrix(mol)
for a in mol.GetAtoms():
idx = a.GetIdx()
if sum(adj_matrix[idx]) > 0:
ave_bond_length = np.mean(dist_matrix[idx][adj_matrix[idx] == 1])
ave_neighbor_wt = np.mean(
[n.GetAtomicNum() for n in a.GetNeighbors()])
else:
ave_bond_length, ave_neighbor_wt = 0.0, 0.0
sym = a.GetSymbol()
a_feats = one_hot_encoding(sym, C.SYMBOLS) \
+ one_hot_encoding(a.GetDegree(), C.DEGREES) \
+ one_hot_encoding(a.GetHybridization(), C.HYBRIDIZATIONS) \
+ [a.GetIsAromatic(), a.GetFormalCharge(), a.GetAtomicNum(),
ave_bond_length, ave_neighbor_wt]
features[idx, :len(a_feats)] = np.array(a_feats)
feat_factory = ChemicalFeatures.BuildFeatureFactory(C.FDEF)
try:
chem_feats = feat_factory.GetFeaturesForMol(mol)
for t in range(len(chem_feats)):
if chem_feats[t].GetFamily() == 'Donor':
for i in chem_feats[t].GetAtomIds():
features[i, -2] = 1
elif chem_feats[t].GetFamily() == 'Acceptor':
for i in chem_feats[t].GetAtomIds():
features[i, -1] = 1
except RuntimeError as e:
print(e)
return features
def testBasic(self):
cfac = ChemicalFeatures.BuildFeatureFactory(
os.path.join(RDConfig.RDBaseDir, 'Code', 'GraphMol',
'MolChemicalFeatures', 'test_data', 'featDef.txt'))
self.failUnless(cfac.GetNumFeatureDefs() == 2)
fNames = cfac.GetFeatureFamilies()
self.failUnless(len(fNames) == 2)
self.failUnless(fNames[0] == 'HBondDonor')
self.failUnless(fNames[1] == 'HBondAcceptor')
mol = Chem.MolFromSmiles("COCN")
rdDistGeom.EmbedMolecule(mol,
30,
100,
useExpTorsionAnglePrefs=False,
useBasicKnowledge=False)
self.failUnless(cfac.GetNumMolFeatures(mol) == 3)
for i in range(cfac.GetNumMolFeatures(mol)):
self.failUnless(cfac.GetMolFeature(mol, i))
# check that the recompute argument works:
self.failUnless(cfac.GetMolFeature(mol, 0))
for i in range(cfac.GetNumMolFeatures(mol)):
self.failUnless(cfac.GetMolFeature(mol, i, "", False))
self.failUnlessRaises(IndexError, lambda: cfac.GetMolFeature(mol, 3))
feats = cfac.GetFeaturesForMol(mol)
self.failUnless(len(feats) == 3)
fTypes = ['HBondDonor', 'HBondAcceptor', 'HBondAcceptor']
positions = [[1.3041, -0.6079, 0.0924], [-0.7066, 0.5994, 0.1824],
[1.3041, -0.6079, 0.0924]]
targetAids = [[3], [1], [3]]
for i, feat in enumerate(feats):
self.assertEqual(feat.GetFamily(), fTypes[i])
pos = list(feat.GetPos())
aids = list(feat.GetAtomIds())
self.assertEqual(aids, targetAids[i])
self.assertTrue(lstFeq(pos, positions[i]))
nmol = feat.GetMol()
self.assertEqual(Chem.MolToSmiles(nmol), "COCN")
ncfac = feat.GetFactory()
self.assertEqual(ncfac.GetNumFeatureDefs(), 2)
self.assertEqual(feat.GetActiveConformer(), -1)
def get_node_features(self, mol):
fdef_name = osp.join(RDConfig.RDDataDir, 'BaseFeatures.fdef')
factory = ChemicalFeatures.BuildFeatureFactory(fdef_name)
x_one_hot = []
x_normal = []
donor = []
acceptor = []
for i in range(2):
x_one_hot.append([])
for i in range(11):
x_normal.append([])
for atom in mol.GetAtoms():
x_one_hot[0].append(atom.GetAtomicNum())
x_one_hot[1].append(atom.GetHybridization())
for atom in mol.GetAtoms():
donor.append(0)
acceptor.append(0)
x_normal[0].append(atom.GetTotalValence())
x_normal[1].append(atom.GetNumImplicitHs())
x_normal[2].append(atom.GetFormalCharge())
x_normal[3].append(atom.GetNumRadicalElectrons())
x_normal[4].append(atom.GetImplicitValence())
x_normal[5].append(atom.GetNumExplicitHs())
x_normal[6].append(atom.GetIsAromatic())
x_normal[7].append(atom.GetIsotope())
x_normal[8].append(atom.GetChiralTag())
feats = factory.GetFeaturesForMol(mol)
for j in range(0, len(feats)):
if feats[j].GetFamily() == 'Donor':
node_list = feats[j].GetAtomIds()
for k in node_list:
donor[k] = 1
elif feats[j].GetFamily() == 'Acceptor':
node_list = feats[j].GetAtomIds()
for k in node_list:
acceptor[k] = 1
x_normal[9] = donor
x_normal[10] = acceptor
x_one_hot = np.array(x_one_hot).T
x_normal = np.array(x_normal).T
#x = torch.tensor(x, dtype=torch.float)
return (x_one_hot, x_normal)
def DefaultSigFactory(fdefFile=None,
minPointCount=2,maxPointCount=3,
bins=[(2,3),(3,4),(4,5),(5,6),(6,7),(7,8),(8,100)]):
import SigFactory
from rdkit.Chem import ChemicalFeatures
if fdefFile is None:
from rdkit import RDConfig
import os.path
fdefFile = os.path.join(RDConfig.RDDataDir,'BaseFeatures.fdef')
featFactory = ChemicalFeatures.BuildFeatureFactory(fdefFile,)
factory = SigFactory.SigFactory(featFactory,
skipFeats=('ZnBinder','LumpedHydrophobe'),
minPointCount=minPointCount,
maxPointCount=maxPointCount,
trianglePruneBins=False)
factory.SetBins(tuple(bins))
factory.Init()
return factory