def _getPropertiesRDkit(smallmol):
"""
Returns ndarray of shape (n_atoms x n_properties) molecule atom types,
according to the following definitions and order:
0. Hydrophibic
1. Aromatic
2. Acceptor
3. Donor
4. - Ionizable
5. + Ionizable
6. Metal (empty)
7. Occupancy (No hydrogens)
"""
from moleculekit.smallmol.util import factory
n_atoms = smallmol.numAtoms
atom_mapping = {
"Hydrophobe": 0,
"LumpedHydrophobe": 0,
"Aromatic": 1,
"Acceptor": 2,
"Donor": 3,
"PosIonizable": 4,
"NegIonizable": 5,
}
feats = factory.GetFeaturesForMol(smallmol._mol)
properties = np.zeros((n_atoms, 8), dtype=bool)
for feat in feats:
fam = feat.GetFamily()
if fam not in atom_mapping: # Non relevant property
continue
properties[feat.GetAtomIds(), atom_mapping[fam]] = 1
# Occupancy, ignoring hydrogens.
properties[:, 7] = smallmol.get("element") != "H"
return properties
def _getPropertiesRDkit(smallmol):
n_atoms = smallmol.numAtoms
from tempfile import NamedTemporaryFile
tmpmol2 = NamedTemporaryFile(suffix='.pdb').name
smallmol.write(tmpmol2)
s_atoms = process_arpeggio(tmpmol2)
os.remove(tmpmol2)
# the order of s_atoms matches the order of channels
arpeggio_features = []
for i, atom in enumerate(s_atoms):
raw_features = sorted(tuple(s_atoms[i].atom_types))
feature_vector = np.zeros(len(arpeggio_atomtypes_unique), dtype=bool)
for rf in raw_features:
#print(rf)
if rf in arpeggio_atomtypes_unique:
feature_vector[arpeggio_atomtypes_unique[rf]] = 1
arpeggio_features.append(feature_vector)
if feature_vector[-1] != 0:
print(raw_features, feature_vector)
arpeggio_features = np.stack(arpeggio_features, axis=0)
#atoms = ['C', 'O', 'N', 'S', 'P', 'Cl', 'F']
atom_mapping = {
"Hydrophobe": 0,
"LumpedHydrophobe": 0,
"Aromatic": 1,
"Acceptor": 2,
"Donor": 3,
"PosIonizable": 4,
"NegIonizable": 5,
"ZnBinder": 6,
}
# other features come from arpeggio
atom_mapping1 = {
"Hydrophobe": 0,
"LumpedHydrophobe": 1,
"ZnBinder": 2,
}
feats = factory.GetFeaturesForMol(smallmol._mol)
properties = np.zeros((n_atoms, 3), dtype=bool)
protein_feature = np.zeros(shape=(len(properties), 1),
dtype=bool) # 0 for ligand
for feat in feats:
fam = feat.GetFamily()
if fam not in atom_mapping1: # Non relevant property
continue
properties[feat.GetAtomIds(), atom_mapping1[fam]] = 1
#for idx, atom in enumerate(atoms):
# properties[:, 7+idx] = smallmol.get('element') == atom
els = np.stack([get_atomic_feature(e) for e in smallmol.get('element')],
axis=0)
# Occupancy, ignoring hydrogens.
#properties[:, 3] = ~np.isin(smallmol.get('element'), ['H'])
print([
c.shape for c in [properties, arpeggio_features, els, protein_feature]
])
channels = np.concatenate(
[properties, arpeggio_features, els, protein_feature], axis=-1)
return channels