def test_at_least_something_matches_every_negative_example(self):
negative_examples = []
substructs = []
for row in GetHeterocycleReactionSmarts():
for smiles in row.NEGATIVE_EXAMPLE.split(','):
if not smiles:
continue
mol = Chem.MolFromSmiles(smiles)
assert mol is not None
negative_examples.append(mol)
smarts = row.SMARTS
if not smarts:
continue
substruct = Chem.MolFromSmarts(smarts)
assert substruct is not None
substructs.append(substruct)
for mol in negative_examples:
something_hit = False
for substruct in substructs:
if mol.HasSubstructMatch(substruct):
something_hit = True
break
assert something_hit, ('nothing matched %s' %
Chem.MolToSmiles(mol, isomericSmiles=True))
def get_six_member_ring_carbon_to_nitrogen_reaction(self):
expected_description = 'aromatic carbon in 6 membered ring'
rxns = [
r for r in GetHeterocycleReactionSmarts()
if r.DESCRIPTION.startswith(expected_description)
]
assert len(rxns) == 1, "expecting only one of these rules for now"
return rxns[0]
def test_reactions_modify_examples(self):
for row in GetHeterocycleReactionSmarts():
smarts = row.SMARTS
if not smarts:
continue
for product in row.CONVERT_TO.split(','):
reaction = smarts + '>>' + product
rxn = AllChem.ReactionFromSmarts(reaction)
for smiles in row.EXAMPLE.split(','):
orig_can_smi = Chem.CanonSmiles(smiles)
assert smiles
mol = Chem.MolFromSmiles(smiles)
for newmol in rxn.RunReactants((mol, )):
newmol = newmol[0]
isosmi = Chem.MolToSmiles(newmol, isomericSmiles=True)
assert_valid_change(orig_can_smi, isosmi)
def test_smarts_match_examples(self):
for row in GetHeterocycleReactionSmarts():
smarts = row.SMARTS
if not smarts:
continue
substruct = Chem.MolFromSmarts(smarts)
for smiles in row.EXAMPLE.split(','):
assert smiles
mol = Chem.MolFromSmiles(smiles)
assert mol.HasSubstructMatch(
substruct), "%s not in %s" % (smarts, smiles)
for smiles in row.NEGATIVE_EXAMPLE.split(','):
if not smiles:
continue
mol = Chem.MolFromSmiles(smiles)
assert not mol.HasSubstructMatch(
substruct), "%s should not be in %s" % (smarts, smiles)
def test_apply_every_rule_to_every_fragment(self):
fieldnames = ['SMILES', 'MUTATED', 'REACTION', 'DESCRIPTION']
writer = csv.DictWriter(open('hetero_atom_mutations.csv', 'w'),
fieldnames)
writer.writeheader()
notchanged = csv.DictWriter(open('not_changed.csv', 'w'),
['SMILES', 'TITLE'])
notchanged.writeheader()
fragment_library = os.path.join(os.path.dirname(__file__), 'test_data',
'fragments.csv')
frag_reader = csv.DictReader(open(fragment_library))
for row in frag_reader:
smiles = row['SMILES']
rdkit_mol = Chem.MolFromSmiles(smiles)
orig_can_smi = Chem.MolToSmiles(rdkit_mol, isomericSmiles=True)
changed = False
for src, rxn in zip(GetHeterocycleReactionSmarts(),
GetHeterocycleReactions()):
for smiles in get_unique_products(rxn, rdkit_mol):
assert_valid_change(orig_can_smi, smiles)
row = {
'SMILES': orig_can_smi,
'MUTATED': smiles,
'REACTION': src.SMARTS + '>>' + src.CONVERT_TO,
'DESCRIPTION': src.DESCRIPTION,
}
writer.writerow(row)
changed = True
# record aromatic fragments that no rule changes (possible problems?)
if not changed and has_aromatic(rdkit_mol):
row = {'SMILES': orig_can_smi, 'TITLE': orig_can_smi}
notchanged.writerow(row)
def test_fuzz_atom_mutations(self):
fragment_library = os.path.join(os.path.dirname(__file__), 'test_data',
'fragments.csv')
base, ext = os.path.splitext(os.path.basename(fragment_library))
rand = Random(0xDEADBEEF)
uniq_fragments = set()
fragments = []
frag_reader = csv.DictReader(open(fragment_library))
for row in frag_reader:
smiles = row['SMILES']
rdkit_mol = Chem.MolFromSmiles(smiles)
if not has_aromatic(rdkit_mol):
continue
orig_can_smi = Chem.MolToSmiles(rdkit_mol, isomericSmiles=True)
assert orig_can_smi not in uniq_fragments
uniq_fragments.add(orig_can_smi)
fragments.append(orig_can_smi)
print(len(fragments), "fragments with aromaticity")
fieldnames = ['SMILES', 'MUTATED', 'REACTION', 'DESCRIPTION']
writer = csv.DictWriter(
open(base + 'hetero_atom_mutations_fuzzing.csv', 'w'), fieldnames)
writer.writeheader()
notchanged = csv.DictWriter(
open(base + 'not_changed_during_fuzzing.csv', 'w'),
['SMILES', 'TITLE'])
notchanged.writeheader()
uniq_notchanged = set()
num_trials = 1000
# to test the full range of possible fragments
#num_trials = 1000000
total_generated = 0
for i in range(num_trials):
if i and i % 1000 == 0:
print(i)
if not fragments:
print("Converged! No more fragments left!")
break
idx = rand.randint(0, len(fragments) - 1)
orig_can_smi = fragments.pop(idx)
rdkit_mol = Chem.MolFromSmiles(orig_can_smi)
changed = False
for src, rxn in zip(GetHeterocycleReactionSmarts(),
GetHeterocycleReactions()):
for smiles in get_unique_products(rxn, rdkit_mol):
total_generated += 1
changed = True
assert_valid_change(orig_can_smi, smiles)
if smiles in uniq_fragments:
continue
uniq_fragments.add(smiles)
fragments.append(smiles)
row = {
'SMILES': orig_can_smi,
'MUTATED': smiles,
'REACTION': src.SMARTS + '>>' + src.CONVERT_TO,
'DESCRIPTION': src.DESCRIPTION,
}
writer.writerow(row)
# record aromatic fragments that no rule changes (possible problems?)
if not changed and orig_can_smi not in uniq_notchanged:
uniq_notchanged.add(orig_can_smi)
row = {'SMILES': orig_can_smi, 'TITLE': orig_can_smi}
notchanged.writerow(row)
print(total_generated, "generated of which", len(uniq_fragments),
"are unique fragments generated after", num_trials, "trials")