def from_molecules(self, mols: Molecules):
tox_data = mols.get_mols_with_passfail_labels()
X = tox_data.index
y = tox_data.astype('int')
self.language.fit(mols.get_all_mols(), X, y)
self.toxicity = Toxicity(self.hyperparams, self.language)
self.toxicity.build(X, y)
print(f"Trained new chemistry model: {self.uuid}")
def to_sql(self, sess: Session):
source_id = add_dataset(sess, self.origin)
add_props(sess, self.properties)
mols = Molecules(sess)
for row in tqdm(self._generate(),
total=len(new_client.molecule),
unit=' row'):
smiles, props = row
mols.add(source_id, smiles, props)
def run_train_models(dbstring: str):
sess = _create_sess(dbstring)
mols = Molecules(sess)
lm = Language()
tox_data = mols.get_mols_with_passfail_labels()
X = tox_data.index
y = tox_data.astype('int')
lm.fit(mols.get_all_mols(), X, y)
Path('data/models').mkdir(parents=True, exist_ok=True)
lm.save('data/models/language.model')
tox = Toxicity(lm)
tox.fit(X, y)
tox.save('data/models/toxicity.model')
def to_sql(self, sess: Session):
source_id = add_dataset(sess, self.origin)
add_props(sess, self.props)
mols = Molecules(sess)
mols.add(source_id, 'CN1CCC[C@H]1c2cccnc2', {'Tag': 'Test'},
PartitionCategory.Unspecific)
mols.add(source_id,
'O1C=C[C@H]([C@H]1O2)c3c2cc(OC)c4c3OC(=O)C5=C4CCC(=O)5', {
'Tag': ['Test1', 'Test2'],
'NR-AR': 1.0
}, PartitionCategory.Verify)
mols.commit()
def run_train_models(dbstring: str, hp=Hyperparameters()) -> Chemistry:
sess = _create_sess(dbstring)
mols = Molecules(sess)
chem = Chemistry(hp)
chem.from_molecules(mols)
Path('data', 'chemistry').mkdir(parents=True, exist_ok=True)
with open(Path('data', 'chemistry', chem.uuid), 'wb') as fd:
pickle.dump(chem, fd)
return chem
def to_sql(self, sess: Session):
source_id = add_dataset(sess, self.origin)
add_props(sess, self.properties)
mols = Molecules(sess)
df = self.to_df()
row_count, _ = df.shape
for _, row in tqdm(df.iterrows(), total=row_count, unit=' row'):
mols.add(source_id, row.smiles, row.to_dict())
mols.commit()
def to_sql(self, sess: Session):
source_id = add_dataset(sess, self.origin)
mols = Molecules(sess)
df = self.to_df()
row_count, _ = df.shape
for _, row in tqdm(df.iterrows(), total=row_count, unit=' row'):
if row.SPLIT == 'train':
partition = PartitionCategory.Train
elif row.SPLIT == 'test':
partition = PartitionCategory.Test
else:
partition = PartitionCategory.Unspecific
mols.add(source_id, row.SMILES, {'Tag': 'MOSES'}, partition)
mols.commit()