def remove_invalid(gen, canonize=True, n_jobs=1):
"""
Removes invalid molecules from the dataset
"""
if not canonize:
mols = mapper(n_jobs)(get_mol, gen)
return [gen_ for gen_, mol in zip(gen, mols) if mol is not None]
return [x for x in mapper(n_jobs)(canonic_smiles, gen) if x is not None]
def frechet_distance(ref, gen, func=None, n_jobs=1):
if func is not None:
ref_values = mapper(n_jobs)(func, ref)
gen_values = mapper(n_jobs)(func, gen)
else:
ref_values = ref
gen_values = gen
ref_mean = np.mean(ref_values)
ref_var = np.var(ref_values)
gen_mean = np.mean(gen_values)
gen_var = np.var(gen_values)
return calculate_frechet_distance(ref_mean, ref_var, gen_mean, gen_var)
def get_all_metrics(ref,
gen,
k=[1000, 10000],
n_jobs=1,
gpu=-1,
batch_size=512):
'''
Computes all available metrics between two lists of SMILES:
* %valid
----- Next metrics are only computed for valid molecules -----
* %unique@k
* Frechet ChemNet Distance (FCD)
* fragment similarity
* scaffold similarity
* morgan similarity
'''
metrics = {}
if n_jobs != 1:
pool = Pool(n_jobs)
else:
pool = 1
metrics['valid'] = fraction_valid(gen, n_jobs=n_jobs)
gen = remove_invalid(gen, canonize=True)
ref = remove_invalid(ref, canonize=True)
gen_mols = mapper(pool)(get_mol, gen)
ref_mols = mapper(pool)(get_mol, ref)
if not isinstance(k, (list, tuple)):
k = [k]
for k_ in k:
metrics['unique@{}'.format(k_)] = fraction_unique(gen, k_, pool)
metrics['FCD'] = frechet_chemnet_distance(ref,
gen,
gpu=gpu,
batch_size=batch_size)
metrics['SNN'] = morgan_similarity(ref_mols, gen_mols, pool, gpu=gpu)
metrics['Frag'] = fragment_similarity(ref_mols, gen_mols, pool)
metrics['Scaf'] = scaffold_similarity(ref_mols, gen_mols, pool)
metrics['IntDiv'] = internal_diversity(gen_mols, pool)
metrics['Filters'] = fraction_passes_filters(gen_mols, pool)
metrics['logP'] = frechet_distance(ref_mols, gen_mols, logP, pool)
metrics['SA'] = frechet_distance(ref_mols, gen_mols, SA, pool)
metrics['QED'] = frechet_distance(ref_mols, gen_mols, QED, pool)
metrics['NP'] = frechet_distance(ref_mols, gen_mols, NP, pool)
metrics['weight'] = frechet_distance(ref_mols, gen_mols, weight, pool)
if n_jobs != 1:
pool.close()
return metrics
def _get_metrics(self, ref, ref_mols, rollout):
rollout_mols = mapper(self.n_jobs)(get_mol, rollout)
result = [[0 if m is None else 1] for m in rollout_mols]
if sum([r[0] for r in result], 0) == 0:
return result
rollout = remove_invalid(rollout, canonize=True, n_jobs=self.n_jobs)
rollout_mols = mapper(self.n_jobs)(get_mol, rollout)
if len(rollout) < 2:
return result
if len(self.metrics):
for metric_name in self.metrics:
if metric_name == 'fcd':
m = FCDMetric(n_jobs=self.n_jobs)(ref, rollout)
elif metric_name == 'morgan':
m = SNNMetric(n_jobs=self.n_jobs)(ref_mols, rollout_mols)
elif metric_name == 'fragments':
m = FragMetric(n_jobs=self.n_jobs)(ref_mols, rollout_mols)
elif metric_name == 'scaffolds':
m = ScafMetric(n_jobs=self.n_jobs)(ref_mols, rollout_mols)
elif metric_name == 'internal_diversity':
m = internal_diversity(rollout_mols, n_jobs=self.n_jobs)
elif metric_name == 'filters':
m = fraction_passes_filters(rollout_mols,
n_jobs=self.n_jobs)
elif metric_name == 'logp':
m = -FrechetMetric(func=logP, n_jobs=self.n_jobs)(
ref_mols, rollout_mols)
elif metric_name == 'sa':
m = -FrechetMetric(func=SA, n_jobs=self.n_jobs)(
ref_mols, rollout_mols)
elif metric_name == 'qed':
m = -FrechetMetric(func=QED, n_jobs=self.n_jobs)(
ref_mols, rollout_mols)
elif metric_name == 'np':
m = -FrechetMetric(func=NP, n_jobs=self.n_jobs)(
ref_mols, rollout_mols)
elif metric_name == 'weight':
m = -FrechetMetric(func=weight, n_jobs=self.n_jobs)(
ref_mols, rollout_mols)
m = MetricsReward._nan2zero(m)
for i in range(len(rollout)):
result[i].append(m)
return result
def fraction_valid(gen, n_jobs=1):
'''
Computes a number of valid molecules
:param gen: list of SMILES
'''
gen = mapper(n_jobs)(get_mol, gen)
return 1 - gen.count(None) / len(gen)
def compute_intermediate_statistics(smiles, n_jobs=1, gpu=-1, batch_size=512):
'''
The function precomputes statistics such as mean and variance for FCD, etc.
It is useful to compute the statistics for test and scaffold test sets to
speedup metrics calculation.
'''
if n_jobs != 1:
pool = Pool(n_jobs)
else:
pool = 1
statistics = {}
mols = mapper(pool)(get_mol, smiles)
kwargs = {'n_jobs': n_jobs, 'gpu': gpu, 'batch_size': batch_size}
statistics['FCD'] = FCDMetric(**kwargs).precalc(smiles)
statistics['SNN'] = SNNMetric(**kwargs).precalc(mols)
statistics['Frag'] = FragMetric(**kwargs).precalc(mols)
statistics['Scaf'] = ScafMetric(**kwargs).precalc(mols)
for name, func in [('logP', logP), ('SA', SA),
('QED', QED), ('NP', NP),
('weight', weight)]:
statistics[name] = FrechetMetric(func, **kwargs).precalc(mols)
if n_jobs != 1:
pool.terminate()
return statistics
def compute_intermediate_statistics(smiles, n_jobs=1, device='cpu',
batch_size=512, pool=None):
"""
The function precomputes statistics such as mean and variance for FCD, etc.
It is useful to compute the statistics for test and scaffold test sets to
speedup metrics calculation.
"""
close_pool = False
if pool is None:
if n_jobs != 1:
pool = Pool(n_jobs)
close_pool = True
else:
pool = 1
statistics = {}
mols = mapper(pool)(get_mol, smiles)
kwargs = {'n_jobs': pool, 'device': device, 'batch_size': batch_size}
kwargs_fcd = {'n_jobs': n_jobs, 'device': device, 'batch_size': batch_size}
statistics['FCD'] = FCDMetric(**kwargs_fcd).precalc(smiles)
statistics['SNN'] = SNNMetric(**kwargs).precalc(mols)
statistics['Frag'] = FragMetric(**kwargs).precalc(mols)
statistics['Scaf'] = ScafMetric(**kwargs).precalc(mols)
for name, func in [('logP', logP), ('SA', SA),
('QED', QED), ('NP', NP),
('weight', weight)]:
statistics[name] = FrechetMetric(func, **kwargs).precalc(mols)
if close_pool:
pool.terminate()
return statistics
def fit(self, data):
"""
Collects fragment frequencies in a training set
Arguments:
data: list of SMILES, training dataset
"""
# Split molecules from dataset into BRICS fragments
fragments = mapper(self.n_jobs)(fragmenter, data)
# Compute fragment frequencies
counts = Counter()
for mol_frag in fragments:
counts.update(mol_frag)
counts = pd.DataFrame(pd.Series(counts).items(),
columns=['fragment', 'count'])
counts['attachment_points'] = [
fragment.count('*') for fragment in counts['fragment'].values
]
counts['frequency'] = counts['count'] / counts['count'].sum()
self.fragment_counts = counts
# Compute number of fragments distribution
fragments_count_distribution = Counter([len(f) for f in fragments])
total = sum(fragments_count_distribution.values())
for k in fragments_count_distribution:
fragments_count_distribution[k] /= total
self.fragments_count_distribution = fragments_count_distribution
self.fitted = True
return self
def reproduce(seed,
samples_path=None,
metrics_path=None,
n_jobs=1,
device='cpu',
verbose=False,
samples=30000):
train = moses.get_dataset('train')
model = CombinatorialGenerator(n_jobs=n_jobs)
if verbose:
print("Training...")
model.fit(train)
if verbose:
print(f"Sampling for seed {seed}")
seeds = list(range((seed - 1) * samples, seed * samples))
samples = mapper(n_jobs)(model.generate_one, seeds)
if samples_path is not None:
with open(samples_path, 'w') as f:
f.write('SMILES\n')
for sample in samples:
f.write(sample + '\n')
if verbose:
print(f"Computing metrics for seed {seed}")
metrics = moses.get_all_metrics(samples, n_jobs=n_jobs, device=device)
if metrics_path is not None:
with open(metrics_path, 'w') as f:
for key, value in metrics.items():
f.write("%s,%f\n" % (key, value))
return samples, metrics
def compute_fragments(mol_list, n_jobs=1):
"""
fragment list of mols using BRICS and return smiles list
"""
fragments = Counter()
for mol_frag in mapper(n_jobs)(fragmenter, mol_list):
fragments.update(mol_frag)
return fragments
def generate(self, n, seed=1, mode=0, verbose=False):
self.set_mode(mode)
seeds = range((seed - 1) * n, seed * n)
if verbose:
print('generating...')
seeds = tqdm(seeds, total=n)
samples = mapper(self.n_jobs)(self.generate_one, seeds)
return samples
def get_vocabulary(self, data):
clusters = set()
for mol in tqdm(mapper(self.config.n_jobs)(MolTree, data),
total=len(data),
postfix=['Creating vocab']):
for c in mol.nodes:
clusters.add(c.smiles)
return JTreeVocab(sorted(list(clusters)))
def fraction_valid(gen, n_jobs=1):
"""
Computes a number of valid molecules
Parameters:
gen: list of SMILES
n_jobs: number of threads for calculation
"""
gen = mapper(n_jobs)(get_mol, gen)
return 1 - gen.count(None) / len(gen)
def fraction_passes_filters(gen, n_jobs=1):
"""
Computes the fraction of molecules that pass filters:
* MCF
* PAINS
* Only allowed atoms ('C','N','S','O','F','Cl','Br','H')
* No charges
"""
passes = mapper(n_jobs)(mol_passes_filters, gen)
return np.mean(passes)
def __init__(self, ref, n_ref_subsample, n_rollouts, n_jobs, metrics=[]):
assert all([m in MetricsReward.supported_metrics for m in metrics])
self.ref = remove_invalid(ref, canonize=True, n_jobs=n_jobs)
self.ref_mols = mapper(n_jobs)(get_mol, self.ref)
self.n_ref_subsample = n_ref_subsample
self.n_rollouts = n_rollouts
self.n_jobs = n_jobs
self.metrics = metrics
def compute_scaffolds(mol_list, n_jobs=1, min_rings=2):
"""
Extracts a scafold from a molecule in a form of a canonic SMILES
"""
scaffolds = Counter()
map_ = mapper(n_jobs)
scaffolds = Counter(
map_(partial(compute_scaffold, min_rings=min_rings), mol_list))
if None in scaffolds:
scaffolds.pop(None)
return scaffolds
def fingerprints(smiles_mols_array,
n_jobs=1,
already_unique=False,
*args,
**kwargs):
'''
Computes fingerprints of smiles np.array/list/pd.Series with n_jobs workers.
e.g.fingerprints(smiles_mols_array, type='morgan', n_jobs=10)
Inserts np.NaN to rows corresponding to incorrect smiles.
IMPORTANT: if there is at least one np.NaN, the dtype would be float
:param smiles_mols_array: list/array/pd.Series of smiles or already computed RDKit molecules
:param n_jobs: number of parralel workers to execute
:param already_unique: flag for performance reasons, if smiles array is big and already unique.
Its value is set to True if smiles_mols_array contain RDKit molecules already.
'''
if isinstance(smiles_mols_array, pd.Series):
smiles_mols_array = smiles_mols_array.values
else:
smiles_mols_array = np.asarray(smiles_mols_array)
if not isinstance(smiles_mols_array[0], str):
already_unique = True
if not already_unique:
smiles_mols_array, inv_index = np.unique(smiles_mols_array,
return_inverse=True)
fps = mapper(n_jobs)(partial(fingerprint, *args, **kwargs),
smiles_mols_array)
length = 1 # Need to know the length to convert None into np.array with nan values
for fp in fps:
if fp is not None:
length = fp.shape[-1]
first_fp = fp
break
fps = [
fp if fp is not None else np.array([np.NaN]).repeat(length)[None, :]
for fp in fps
]
if scipy.sparse.issparse(first_fp):
fps = scipy.sparse.vstack(fps).tocsr()
else:
fps = np.vstack(fps)
if not already_unique:
return fps[inv_index]
else:
return fps
def fraction_unique(gen, k=None, n_jobs=1, check_validity=True):
'''
Computes a number of unique molecules
:param gen: list of SMILES
:param k: compute unique@k
:param check_validity: raises ValueError if invalid molecules are present
'''
if k is not None:
if len(gen) < k:
warnings.warn(
"Can't compute unique@{}. gen contains only {} molecules".
format(k, len(gen)))
gen = gen[:k]
canonic = set(mapper(n_jobs)(canonic_smiles, gen))
if None in canonic and check_validity:
raise ValueError("Invalid molecule passed to unique@k")
return len(canonic) / len(gen)
def fraction_unique(gen, k=None, n_jobs=1, check_validity=True):
"""
Computes a number of unique molecules
Parameters:
gen: list of SMILES
k: compute unique@k
n_jobs: number of threads for calculation
check_validity: raises ValueError if invalid molecules are present
"""
if k is not None:
if len(gen) < k:
warnings.warn("Can't compute unique@{}.".format(k) +
"gen contains only {} molecules".format(len(gen)))
gen = gen[:k]
canonic = set(mapper(n_jobs)(canonic_smiles, gen))
if None in canonic and check_validity:
raise ValueError("Invalid molecule passed to unique@k")
return len(canonic) / len(gen)
def precalc(self, mols):
if self.func is not None:
values = mapper(self.n_jobs)(self.func, mols)
else:
values = mols
return {'mu': np.mean(values), 'var': np.var(values)}
def novelty(gen, train, n_jobs=1):
gen_smiles = mapper(n_jobs)(canonic_smiles, gen)
gen_smiles_set = set(gen_smiles) - {None}
train_set = set(train)
return len(gen_smiles_set - train_set) / len(gen_smiles_set)
def get_all_metrics(test, gen, k=None, n_jobs=1, device='cpu',
batch_size=512, test_scaffolds=None,
ptest=None, ptest_scaffolds=None,
pool=None, gpu=None, train=None):
"""
Computes all available metrics between test (scaffold test)
and generated sets of SMILES.
Parameters:
test: list of test SMILES
gen: list of generated SMILES
k: int or list with values for unique@k. Will calculate number of
unique molecules in the first k molecules. Default [1000, 10000]
n_jobs: number of workers for parallel processing
device: 'cpu' or 'cuda:n', where n is GPU device number
batch_size: batch size for FCD metric
test_scaffolds: list of scaffold test SMILES
Will compute only on the general test set if not specified
ptest: dict with precalculated statistics of the test set
ptest_scaffolds: dict with precalculated statistics
of the scaffold test set
pool: optional multiprocessing pool to use for parallelization
gpu: deprecated, use `device`
train: list of train SMILES
Available metrics:
* %valid
* %unique@k
* Frechet ChemNet Distance (FCD)
* Fragment similarity (Frag)
* Scaffold similarity (Scaf)
* Similarity to nearest neighbour (SNN)
* Internal diversity (IntDiv)
* Internal diversity 2: using square root of mean squared
Tanimoto similarity (IntDiv2)
* %passes filters (Filters)
* Distribution difference for logP, SA, QED, NP, weight
* Novelty (molecules not present in train)
"""
if k is None:
k = [1000, 10000]
disable_rdkit_log()
metrics = {}
if gpu is not None:
warnings.warn(
"parameter `gpu` is deprecated. Use `device`",
DeprecationWarning
)
if gpu == -1:
device = 'cpu'
else:
device = 'cuda:{}'.format(gpu)
close_pool = False
if pool is None:
if n_jobs != 1:
pool = Pool(n_jobs)
close_pool = True
else:
pool = 1
metrics['valid'] = fraction_valid(gen, n_jobs=pool)
gen = remove_invalid(gen, canonize=True)
if not isinstance(k, (list, tuple)):
k = [k]
for _k in k:
metrics['unique@{}'.format(_k)] = fraction_unique(gen, _k, pool)
if ptest is None:
ptest = compute_intermediate_statistics(test, n_jobs=n_jobs,
device=device,
batch_size=batch_size,
pool=pool)
if test_scaffolds is not None and ptest_scaffolds is None:
ptest_scaffolds = compute_intermediate_statistics(
test_scaffolds, n_jobs=n_jobs,
device=device, batch_size=batch_size,
pool=pool
)
mols = mapper(pool)(get_mol, gen)
kwargs = {'n_jobs': pool, 'device': device, 'batch_size': batch_size}
kwargs_fcd = {'n_jobs': n_jobs, 'device': device, 'batch_size': batch_size}
metrics['FCD/Test'] = FCDMetric(**kwargs_fcd)(gen=gen, pref=ptest['FCD'])
metrics['SNN/Test'] = SNNMetric(**kwargs)(gen=mols, pref=ptest['SNN'])
metrics['Frag/Test'] = FragMetric(**kwargs)(gen=mols, pref=ptest['Frag'])
metrics['Scaf/Test'] = ScafMetric(**kwargs)(gen=mols, pref=ptest['Scaf'])
if ptest_scaffolds is not None:
metrics['FCD/TestSF'] = FCDMetric(**kwargs_fcd)(
gen=gen, pref=ptest_scaffolds['FCD']
)
metrics['SNN/TestSF'] = SNNMetric(**kwargs)(
gen=mols, pref=ptest_scaffolds['SNN']
)
metrics['Frag/TestSF'] = FragMetric(**kwargs)(
gen=mols, pref=ptest_scaffolds['Frag']
)
metrics['Scaf/TestSF'] = ScafMetric(**kwargs)(
gen=mols, pref=ptest_scaffolds['Scaf']
)
metrics['IntDiv'] = internal_diversity(mols, pool, device=device)
metrics['IntDiv2'] = internal_diversity(mols, pool, device=device, p=2)
metrics['Filters'] = fraction_passes_filters(mols, pool)
# Properties
for name, func in [('logP', logP), ('SA', SA),
('QED', QED), ('NP', NP),
('weight', weight)]:
metrics[name] = FrechetMetric(func, **kwargs)(gen=mols,
pref=ptest[name])
if train is not None:
metrics['Novelty'] = novelty(mols, train, pool)
enable_rdkit_log()
if close_pool:
pool.close()
pool.join()
return metrics
def precalc(self, mols):
if self.func is not None:
values = mapper(self.n_jobs)(self.func, mols)
else:
values = mols
return {'values': values}
def get_all_metrics(gen,
k=None,
n_jobs=1,
device='cpu',
batch_size=512,
pool=None,
test=None,
test_scaffolds=None,
ptest=None,
ptest_scaffolds=None,
train=None):
"""
Computes all available metrics between test (scaffold test)
and generated sets of SMILES.
Parameters:
gen: list of generated SMILES
k: int or list with values for unique@k. Will calculate number of
unique molecules in the first k molecules. Default [1000, 10000]
n_jobs: number of workers for parallel processing
device: 'cpu' or 'cuda:n', where n is GPU device number
batch_size: batch size for FCD metric
pool: optional multiprocessing pool to use for parallelization
test (None or list): test SMILES. If None, will load
a default test set
test_scaffolds (None or list): scaffold test SMILES. If None, will
load a default scaffold test set
ptest (None or dict): precalculated statistics of the test set. If
None, will load default test statistics. If you specified a custom
test set, default test statistics will be ignored
ptest_scaffolds (None or dict): precalculated statistics of the
scaffold test set If None, will load default scaffold test
statistics. If you specified a custom test set, default test
statistics will be ignored
train (None or list): train SMILES. If None, will load a default
train set
Available metrics:
* %valid
* %unique@k
* Frechet ChemNet Distance (FCD)
* Fragment similarity (Frag)
* Scaffold similarity (Scaf)
* Similarity to nearest neighbour (SNN)
* Internal diversity (IntDiv)
* Internal diversity 2: using square root of mean squared
Tanimoto similarity (IntDiv2)
* %passes filters (Filters)
* Distribution difference for logP, SA, QED, weight
* Novelty (molecules not present in train)
"""
if test is None:
if ptest is not None:
raise ValueError("You cannot specify custom test "
"statistics for default test set")
test = get_dataset('test')
ptest = get_statistics('test')
if test_scaffolds is None:
if ptest_scaffolds is not None:
raise ValueError("You cannot specify custom scaffold test "
"statistics for default scaffold test set")
test_scaffolds = get_dataset('test_scaffolds')
ptest_scaffolds = get_statistics('test_scaffolds')
train = train or get_dataset('train')
if k is None:
k = [1000, 10000]
disable_rdkit_log()
metrics = {}
close_pool = False
if pool is None:
if n_jobs != 1:
pool = Pool(n_jobs)
close_pool = True
else:
pool = 1
metrics['valid'] = fraction_valid(gen, n_jobs=pool)
gen = remove_invalid(gen, canonize=True)
if not isinstance(k, (list, tuple)):
k = [k]
for _k in k:
metrics['unique@{}'.format(_k)] = fraction_unique(gen, _k, pool)
if ptest is None:
ptest = compute_intermediate_statistics(test,
n_jobs=n_jobs,
device=device,
batch_size=batch_size,
pool=pool)
if test_scaffolds is not None and ptest_scaffolds is None:
ptest_scaffolds = compute_intermediate_statistics(
test_scaffolds,
n_jobs=n_jobs,
device=device,
batch_size=batch_size,
pool=pool)
mols = mapper(pool)(get_mol, gen)
kwargs = {'n_jobs': pool, 'device': device, 'batch_size': batch_size}
kwargs_fcd = {'n_jobs': n_jobs, 'device': device, 'batch_size': batch_size}
metrics['FCD/Test'] = FCDMetric(**kwargs_fcd)(gen=gen, pref=ptest['FCD'])
metrics['SNN/Test'] = SNNMetric(**kwargs)(gen=mols, pref=ptest['SNN'])
metrics['Frag/Test'] = FragMetric(**kwargs)(gen=mols, pref=ptest['Frag'])
metrics['Scaf/Test'] = ScafMetric(**kwargs)(gen=mols, pref=ptest['Scaf'])
if ptest_scaffolds is not None:
metrics['FCD/TestSF'] = FCDMetric(**kwargs_fcd)(
gen=gen, pref=ptest_scaffolds['FCD'])
metrics['SNN/TestSF'] = SNNMetric(**kwargs)(
gen=mols, pref=ptest_scaffolds['SNN'])
metrics['Frag/TestSF'] = FragMetric(**kwargs)(
gen=mols, pref=ptest_scaffolds['Frag'])
metrics['Scaf/TestSF'] = ScafMetric(**kwargs)(
gen=mols, pref=ptest_scaffolds['Scaf'])
metrics['IntDiv'] = internal_diversity(mols, pool, device=device)
metrics['IntDiv2'] = internal_diversity(mols, pool, device=device, p=2)
metrics['Filters'] = fraction_passes_filters(mols, pool)
# Properties
for name, func in [('logP', logP), ('SA', SA), ('QED', QED),
('weight', weight)]:
metrics[name] = WassersteinMetric(func, **kwargs)(gen=mols,
pref=ptest[name])
if train is not None:
metrics['Novelty'] = novelty(mols, train, pool)
enable_rdkit_log()
if close_pool:
pool.close()
pool.join()
return metrics
def get_reference_data(self, data):
ref_smiles = remove_invalid(data, canonize=True, n_jobs=self.n_jobs)
ref_mols = mapper(self.n_jobs)(get_mol, ref_smiles)
return ref_smiles, ref_mols
def get_all_metrics(test,
gen,
k=[1000, 10000],
n_jobs=1,
gpu=-1,
batch_size=512,
test_scaffolds=None,
ptest=None,
ptest_scaffolds=None):
'''
Computes all available metrics between test (scaffold test) and generated sets of SMILES.
Parameters:
test: list of test SMILES
gen: list of generated SMILES
k: list with values for unique@k.
Will calculate number of unique molecules in the first k molecules.
n_jobs: number of workers for parallel processing
gpu: index of GPU for FCD metric and internal diversity, -1 means use CPU
batch_size: batch size for FCD metric
test_scaffolds: list of scaffold test SMILES
Will compute only on the general test set if not specified
ptest: dict with precalculated statistics of the test set
ptest_scaffolds: dict with precalculated statistics of the scaffold test set
Available metrics:
* %valid
* %unique@k
* Frechet ChemNet Distance (FCD)
* Fragment similarity (Frag)
* Scaffold similarity (Scaf)
* Similarity to nearest neighbour (SNN)
* Internal diversity (IntDiv)
* Internal diversity 2: using square root of mean squared Tanimoto similarity (IntDiv2)
* %passes filters (Filters)
* Distribution difference for logP, SA, QED, NP, weight
'''
disable_rdkit_log()
metrics = {}
if n_jobs != 1:
pool = Pool(n_jobs)
else:
pool = 1
metrics['valid'] = fraction_valid(gen, n_jobs=n_jobs)
gen = remove_invalid(gen, canonize=True)
if not isinstance(k, (list, tuple)):
k = [k]
for _k in k:
metrics['unique@{}'.format(_k)] = fraction_unique(gen, _k, pool)
if ptest is None:
ptest = compute_intermediate_statistics(test,
n_jobs=n_jobs,
gpu=gpu,
batch_size=batch_size)
if test_scaffolds is not None and ptest_scaffolds is None:
ptest_scaffolds = compute_intermediate_statistics(
test_scaffolds, n_jobs=n_jobs, gpu=gpu, batch_size=batch_size)
mols = mapper(pool)(get_mol, gen)
kwargs = {'n_jobs': pool, 'gpu': gpu, 'batch_size': batch_size}
metrics['FCD/Test'] = FCDMetric(**kwargs)(gen=gen, ptest=ptest['FCD'])
metrics['SNN/Test'] = SNNMetric(**kwargs)(gen=mols, ptest=ptest['SNN'])
metrics['Frag/Test'] = FragMetric(**kwargs)(gen=mols, ptest=ptest['Frag'])
metrics['Scaf/Test'] = ScafMetric(**kwargs)(gen=mols, ptest=ptest['Scaf'])
if ptest_scaffolds is not None:
metrics['FCD/TestSF'] = FCDMetric(**kwargs)(
gen=gen, ptest=ptest_scaffolds['FCD'])
metrics['SNN/TestSF'] = SNNMetric(**kwargs)(
gen=mols, ptest=ptest_scaffolds['SNN'])
metrics['Frag/TestSF'] = FragMetric(**kwargs)(
gen=mols, ptest=ptest_scaffolds['Frag'])
metrics['Scaf/TestSF'] = ScafMetric(**kwargs)(
gen=mols, ptest=ptest_scaffolds['Scaf'])
metrics['IntDiv'] = internal_diversity(mols, pool, gpu=gpu)
metrics['IntDiv2'] = internal_diversity(mols, pool, gpu=gpu, p=2)
metrics['Filters'] = fraction_passes_filters(mols, pool)
# Properties
for name, func in [('logP', logP), ('SA', SA), ('QED', QED), ('NP', NP),
('weight', weight)]:
metrics[name] = FrechetMetric(func, **kwargs)(gen=mols,
ptest=ptest[name])
enable_rdkit_log()
if n_jobs != 1:
pool.terminate()
return metrics
