def check_data(self, df):
startTime = datetime.datetime.now()
def fnDisplay(message):
display(message, startTime)
smiles = df.smiles
for i in range(len(smiles)):
ion = smiles[i]
try:
Chem.SanitizeMol(Chem.MolFromSmiles(ion))
except ValueError:
name = salty.check_name(ion)
message = "RDKit cannot interpret %s ion SMILES in datafile" \
% name
fnDisplay(message)
if "-" not in ion and "+" not in ion:
name = salty.check_name(ion)
message = "%s ion does not have a charge" % name
fnDisplay(message)
if "." in ion:
name = salty.check_name(ion)
message = "%s ion contains more than one molecular entity" \
% name
fnDisplay(message)
def _show_ion(genes,
target,
mutation_attempts,
sim_score,
molecular_relative,
models,
deslists,
anion_smiles,
exp_data=None):
"""
for printing results to the screen. _show_ion is called when a candidate
has achieved the desired fitness core and is returned by the engine
"""
mol = Chem.MolFromSmiles(genes)
anion = Chem.MolFromSmiles(anion_smiles)
fitness, mol_property = _get_fitness(anion, genes, target, models,
deslists)
anion_name = salty.check_name(anion_smiles)
if exp_data:
chrom = genetic.Chromosome(genes, fitness)
exp_parent_candidates = eval(exp_data.Data_summary.iloc[1][0])
tan_sim_score, sim_index = \
genetic.molecular_similarity(chrom, exp_parent_candidates)
molecular_relative = exp_parent_candidates[sim_index]
print("{}\t{}".format("Salt Smiles: ", genes))
print("{}\t{}".format("Cation Heavy Atoms: ", mol.GetNumAtoms()))
print("Tanimoto Similarity Score: \t{0:10.3f}".format(sim_score))
print("{}\t{}".format("Molecular Relative: ",
salty.check_name(molecular_relative)))
print("{}\t{}".format("Anion: ", anion_name))
print("{}\t{}".format("Model Prediction: ", mol_property))
print("{}\t{}".format("Mutation Attempts: ", mutation_attempts))
def _show_ion(genes, target, mutation_attempts, sim_score, molecular_relative,
models, deslists, anion_smiles):
"""
for printing results to the screen. _show_ion is called when a candidate
has achieved the desired fitness core and is returned by the engine
"""
mol = Chem.MolFromSmiles(genes)
anion = Chem.MolFromSmiles(anion_smiles)
fitness, mol_property = _get_fitness(anion, genes, target,
models, deslists)
anion_name = salty.check_name(anion_smiles)
print("{}\t{}".format("Salt Smiles: ", genes))
print("{}\t{}".format("Cation Heavy Atoms: ", mol.GetNumAtoms()))
print("Tanimoto Similarity Score: \t{0:10.3f}".format(sim_score))
print("{}\t{}".format("Molecular Relative: ",
salty.check_name(molecular_relative)))
print("{}\t{}".format("Anion: ", anion_name))
print("{}\t{}".format("Model Prediction: ", mol_property))
print("{}\t{}".format("Mutation Attempts: ", mutation_attempts))
def _show_ion(genes, target, mutation_attempts, sim_score, molecular_relative,
model_ID, anion):
"""
for printing results to the screen. _show_ion is called when a candidate
has achieved the desired fitness core and is returned by the engine
"""
mol = Chem.MolFromSmiles(genes)
fitness, mol_property = _get_fitness(anion, genes, target, model_ID)
print("{}\t{}".format("number of atoms: ", mol.GetNumAtoms()))
print("{}\t{}".format("mutation attempts: ", mutation_attempts))
print("with prediction: \t{}".format(mol_property))
print("similarity score: {0:10.3f}".format(sim_score))
print("{}\t{}\n".format("molecular relative: ",
salty.check_name(molecular_relative)))
class iupac_smiles_tests(unittest.TestCase):
data_files = ["cationInfo.csv", "anionInfo.csv"]
df = salty.load_data(data_files[0])
smiles = df.smiles
for i in range(len(smiles)):
ion = smiles[i]
salty.check_name(ion)
def test_1_check_data(self):
for i in range(len(self.data_files)):
df = salty.load_data(self.data_files[i])
self.check_data(df)
def test_2_check_wrong_ion(selfs):
ion = 'stupid_nonsense_string'
salty.check_name(ion)
def test_benchmark(self):
salty.Benchmark.run(self.test_1_check_data)
salty.Benchmark.run(self.test_2_check_wrong_ion)
def check_data(self, df):
startTime = datetime.datetime.now()
def fnDisplay(message):
display(message, startTime)
smiles = df.smiles
for i in range(len(smiles)):
ion = smiles[i]
try:
Chem.SanitizeMol(Chem.MolFromSmiles(ion))
except ValueError:
name = salty.check_name(ion)
message = "RDKit cannot interpret %s ion SMILES in datafile" \
% name
fnDisplay(message)
if "-" not in ion and "+" not in ion:
name = salty.check_name(ion)
message = "%s ion does not have a charge" % name
fnDisplay(message)
if "." in ion:
name = salty.check_name(ion)
message = "%s ion contains more than one molecular entity" \
% name
fnDisplay(message)
def test_2_check_wrong_ion(selfs):
ion = 'stupid_nonsense_string'
salty.check_name(ion)
print('There are ' + str(len(four)) + ' salts of 2 each')
print('There are ' + str(len(more)) + ' salts of 2 or more each')
cation2 = []
anion2 = []
error2_anion = []
error2_cation = []
for i in two:
cation2.append(i[0])
anion2.append(i[1])
for i in anion2: #CHECK CHECK_NAME FUNC FOR MISSING ANION OR CATION
#print(i)
try:
check_name(i)
except:
UnboundLocalError
error2_anion.append(i)
#print(i)
for i in cation2:
#print(i)
try:
check_name(i)
except:
UnboundLocalError
error2_cation.append(i)
print('There are ' + str(len(set(error2_anion))) +
' unique missing anions from the data base')
#print(error2_anion)
def generate_solvent(target,
model_ID,
heavy_atom_limit=50,
sim_bounds=[0.4, 1.0],
hits=1,
write_file=False):
"""
the primary public function of the salt_generator module
Parameters
----------
target : array, float, or int
the desired property value to be achieved by the engine, if
an array, a multi-output model must be supplied to the engine
model_ID : str
the name of the model to be used by the engine. Gains has
several built-in models to choose from
heavy_atom_limit : int, optional
the upper value for allowable heavy atoms in the returned
candidate
sim_bounds : array, optional
the tanimoto similarity score between the returned candidate
and its closest molecular relative in parent_candidates
hits : int, optional
the number of desired solutions
write_file : boolean, optional
defaults to False. if True will return the solutions and a
csv log file
Returns
-------
new : object
default behavior is to return a pandas DataFrame. This is
a log file of the solution(s). if write_file = True the
function will also return pdb files of the cations/anions
"""
parent_candidates = eval(
genetic.load_data("{}_summary.csv".format(model_ID)).loc[1][1])
anion_candidates = eval(
genetic.load_data("{}_summary.csv".format(model_ID)).loc[2][1])
cols = [
"Salt ID", "Salt Smiles", "Cation Heavy Atoms",
"Tanimoto Similarity Score", "Molecular Relative", "Anion",
"Model Prediction", "MD Calculation", "Error"
]
salts = pd.DataFrame(columns=cols)
for i in range(1, hits + 1):
while True:
anion_smiles = random.sample(list(anion_candidates), 1)[0]
anion = Chem.MolFromSmiles(anion_smiles)
best = _guess_password(target, anion, parent_candidates, model_ID)
tan_sim_score, sim_index =\
genetic.molecular_similarity(best, parent_candidates)
cation_heavy_atoms = best.Mol.GetNumAtoms()
salt_smiles = best.Genes + "." + Chem.MolToSmiles(anion)
if cation_heavy_atoms < heavy_atom_limit and\
tan_sim_score >= sim_bounds[0] and\
tan_sim_score < sim_bounds[1] and\
salt_smiles not in salts["Salt Smiles"]:
scr, pre = _get_fitness(anion, best.Genes, target, model_ID)
if i < 10:
CAT_ID = "C0%s" % i
AN_ID = "A0%s" % i
else:
CAT_ID = "C%s" % i
AN_ID = "A%s" % i
salt_ID = CAT_ID + "_" + AN_ID
molecular_relative = salty.check_name(
parent_candidates[sim_index])
anion_name = salty.check_name(anion_smiles)
new_entry = pd.DataFrame([[
salt_ID, salt_smiles, cation_heavy_atoms, tan_sim_score,
molecular_relative, anion_name, pre
]],
columns=cols[:-2])
try:
cation = Chem.AddHs(best.Mol)
Chem.EmbedMolecule(cation, Chem.ETKDG())
Chem.UFFOptimizeMolecule(cation)
anion = Chem.AddHs(anion)
Chem.EmbedMolecule(anion, Chem.ETKDG())
Chem.UFFOptimizeMolecule(anion)
new = pd.DataFrame(pd.concat([salts, new_entry]),
columns=cols)
except BaseException:
continue
if write_file:
MolToPDBFile(cation, "{}.pdb".format(CAT_ID))
MolToPDBFile(anion, "{}.pdb".format(AN_ID))
break
else:
continue
if write_file:
pd.DataFrame.to_csv(new, path_or_buf="salt_log.csv", index=False)
salts = new
if not write_file:
return new
def generate_solvent(target,
model_ID,
heavy_atom_limit=50,
sim_bounds=[0, 1.0],
hits=1,
write_file=False,
seed=None,
hull=None,
simplex=None,
path=None,
exp_data=None,
verbose=0,
gen_token=False,
hull_bounds=[0, 1],
inner_search=True,
parent_cap=25,
mutation_cap=1000):
"""
the primary public function of the salt_generator module
Parameters
----------
target : array, float, or int
the desired property value to be achieved by the engine, if
an array, a multi-output model must be supplied to the engine
model_ID : str
the name of the model to be used by the engine. Gains has
several built-in models to choose from
heavy_atom_limit : int, optional
the upper value for allowable heavy atoms in the returned
candidate
sim_bounds : array, optional
the tanimoto similarity score between the returned candidate
and its closest molecular relative in parent_candidates
hits : int, optional
the number of desired solutions
write_file : boolean, optional
defaults to False. if True will return the solutions and a
csv log file
seed : int, optional
optional randint seed for unittest consistency
hull : pandas DataFrame, optional
nxm pandas DataFrame to use convex hull search strategy. hull
columns should be the same properties used in the genetic algorithm
fitness test
simplex : array, optional
array to access boundary datapoints in the convex hull. This is used
during target resampling defined by the convex hull/simplex
path : str, optional
absolute path to the qspr model used as the fitness function
exp_data: salty devmodel obj, optional
used during hull target reassignment search strategy. Salty devmodel
object of the original experimental data
verbose : int, optional, default 0
0 : most verbose. Best child, parent/target resampling,
sanitization failure
1 : parent/target resampling, solution metadata, sanitization failure
2 : solution metdata, sanitization failure
3 : target resampling, csv-formatted solution metadata
4 : csv-formatted solution metadata
gen_token : int, str, optional
a string or integer to append to file outputs. Useful in the case of
parallel searches.
hull_bounds : array, optional
if hull and simplex are not none, hull_bounds describes the
proximity convex_search should be to the simplex
inner_search : bool, optional
if hull and simplex are not none, inner_search specifies if
convex_search should return values only within the convex hull
Returns
-------
new : object
default behavior is to return a pandas DataFrame. This is
a log file of the solution(s). if write_file = True the
function will also return pdb files of the cations/anions
"""
parent_candidates = []
anion_candidates = []
models = []
deslists = []
for i, name in enumerate(model_ID):
if path:
model = np.array(
[load_model(join(path, '{}_qspr.h5'.format(name)))])
with open(join(path, '{}_desc.csv'.format(name)),
'rb') as csv_file:
deslist = list([pd.read_csv(csv_file, encoding='latin1')])
with open(join(path, '{}_summ.csv'.format(name)),
'rb') as csv_file:
summary = pd.read_csv(csv_file, encoding='latin1')
else:
model = np.array(
[genetic.load_data("{}_qspr.h5".format(name), h5File=True)])
deslist = list([genetic.load_data("{}_desc.csv".format(name))])
summary = genetic.load_data("{}_summ.csv".format(name))
parents = eval(summary.iloc[1][1])
anions = eval(summary.iloc[2][1])
if i > 0:
parent_candidates = np.concatenate((parents, parent_candidates))
anion_candidates = np.concatenate((anions, anion_candidates))
models = np.concatenate((models, model))
deslists = list([deslists, deslist])
else:
parent_candidates = parents
anion_candidates = anions
models = model
deslists = deslist
cols = [
"Salt ID", "Salt Smiles", "Cation Heavy Atoms",
"Tanimoto Similarity Score", "Molecular Relative", "Anion",
"Model Prediction", "MD Calculation", "Error"
]
salts = pd.DataFrame(columns=cols)
if exp_data:
anion_candidates = eval(exp_data.Data_summary.iloc[2][0])
for i in range(1, hits + 1):
while True:
if seed:
random.seed(seed)
anion_smiles = random.sample(list(anion_candidates), 1)[0]
anion = Chem.MolFromSmiles(anion_smiles)
best = _guess_password(target,
anion_smiles,
parent_candidates,
models,
deslists,
seed=seed,
hull=hull,
simplex=simplex,
exp_data=exp_data,
verbose=verbose,
hull_bounds=hull_bounds,
inner_search=inner_search,
parent_cap=parent_cap,
mutation_cap=mutation_cap)
if exp_data:
exp_parent_candidates = eval(exp_data.Data_summary.iloc[1][0])
tan_sim_score, sim_index = \
genetic.molecular_similarity(best, exp_parent_candidates)
else:
tan_sim_score, sim_index = \
genetic.molecular_similarity(best, parent_candidates)
cation_heavy_atoms = best.Mol.GetNumAtoms()
salt_smiles = best.Genes + "." + Chem.MolToSmiles(anion)
if cation_heavy_atoms < heavy_atom_limit and \
sim_bounds[0] <= tan_sim_score < sim_bounds[1] and\
salt_smiles not in salts["Salt Smiles"]:
scr, pre = _get_fitness(anion, best.Genes, target, models,
deslists)
if i < 10:
CAT_ID = "C0%s" % i
AN_ID = "A0%s" % i
else:
CAT_ID = "C%s" % i
AN_ID = "A%s" % i
salt_ID = CAT_ID + "_" + AN_ID
if exp_data:
molecular_relative = salty.check_name(
exp_parent_candidates[sim_index])
else:
molecular_relative = salty.check_name(
parent_candidates[sim_index])
anion_name = salty.check_name(anion_smiles)
new_entry = pd.DataFrame([[
salt_ID, salt_smiles, cation_heavy_atoms, tan_sim_score,
molecular_relative, anion_name, pre
]],
columns=cols[:-2])
try:
cation = Chem.AddHs(best.Mol)
Chem.EmbedMolecule(cation, Chem.ETKDG())
Chem.UFFOptimizeMolecule(cation)
anion = Chem.AddHs(anion)
Chem.EmbedMolecule(anion, Chem.ETKDG())
Chem.UFFOptimizeMolecule(anion)
new = pd.DataFrame(pd.concat([salts, new_entry]),
columns=cols)
except BaseException:
if verbose == any([0, 1, 2]):
print("molecule not sanitizable")
continue
if write_file:
if verbose == any([3, 4]):
print(new)
if gen_token:
MolToPDBFile(cation,
"{}_{}.pdb".format(gen_token, CAT_ID))
MolToPDBFile(anion,
"{}_{}.pdb".format(gen_token, AN_ID))
else:
MolToPDBFile(cation, "{}.pdb".format(CAT_ID))
MolToPDBFile(anion, "{}.pdb".format(AN_ID))
break
else:
continue
if write_file:
if gen_token:
pd.DataFrame.to_csv(
new,
path_or_buf="{}_salt_log.csv".format(gen_token),
index=False)
else:
pd.DataFrame.to_csv(new,
path_or_buf="salt_log.csv",
index=False)
salts = new
if not write_file:
return new