def k_fold_split(self, dataset, k, directories=None, **kwargs):
"""
Parameters
----------
dataset: `dc.data.Dataset`
Dataset to do a k-fold split
k: int
Number of folds to split `dataset` into.
directories: list[str]
list of length 2*k filepaths to save the result disk-datasets
Returns
-------
list of length k tuples of (train, cv) where `train` and `cv` are both
lists of `Dataset`s.
"""
logger.info("Computing K-fold split")
if directories is None:
directories = [tempfile.mkdtemp() for _ in range(2 * k)]
else:
assert len(directories) == 2 * k
cv_datasets = []
train_ds_base = None
train_datasets = []
# rem_dataset is remaining portion of dataset
if isinstance(dataset, DiskDataset):
rem_dataset = dataset
else:
rem_dataset = DiskDataset.from_numpy(dataset.X, dataset.y,
dataset.w, dataset.ids)
for fold in range(k):
# Note starts as 1/k since fold starts at 0. Ends at 1 since fold goes up
# to k-1.
frac_fold = 1. / (k - fold)
train_dir, cv_dir = directories[2 * fold], directories[2 * fold +
1]
fold_inds, rem_inds, _ = self.split(rem_dataset,
frac_train=frac_fold,
frac_valid=1 - frac_fold,
frac_test=0,
**kwargs)
cv_dataset = rem_dataset.select(fold_inds, select_dir=cv_dir)
cv_datasets.append(cv_dataset)
rem_dataset = rem_dataset.select(rem_inds)
train_ds_to_merge = filter(lambda x: x is not None,
[train_ds_base, rem_dataset])
train_ds_to_merge = filter(lambda x: len(x) > 0, train_ds_to_merge)
train_dataset = DiskDataset.merge(train_ds_to_merge,
merge_dir=train_dir)
train_datasets.append(train_dataset)
update_train_base_merge = filter(lambda x: x is not None,
[train_ds_base, cv_dataset])
train_ds_base = DiskDataset.merge(update_train_base_merge)
return list(zip(train_datasets, cv_datasets))
def load_gpcr(dataset_file,
featurizer='ECFP',
transformers=True,
reload=True,
sep='OnePositiveSplit',
K=5):
# data_dir=os.path.dirname(dataset_file)
save_dir = os.path.join(
os.path.dirname(dataset_file),
'.'.join(os.path.basename(dataset_file).split('.')[:-1]), "ecfp",
"split")
train, valid, test = os.path.join(save_dir, 'train'), os.path.join(
save_dir, 'valid'), os.path.join(save_dir, 'test')
fopen = open(dataset_file, "r")
ss = fopen.readlines()
m = ss[0].strip('\n').split(',')
m.remove('SMILES')
if os.path.isdir(save_dir):
if reload:
dataset, train_dataset, valid_dataset, test_dataset = DiskDataset(
data_dir=save_dir), DiskDataset(data_dir=train), DiskDataset(
data_dir=valid), DiskDataset(data_dir=test)
transformers = [
deepchem.trans.NormalizationTransformer(transform_w=True,
dataset=train_dataset)
]
all_dataset = (dataset, train_dataset, valid_dataset, test_dataset)
return m, all_dataset, transformers
if featurizer == 'ECFP':
featurizer = deepchem.feat.CircularFingerprint(size=1024)
elif featurizer == 'GraphConv':
featurizer = deepchem.feat.ConvMolFeaturizer()
elif featurizer == 'Weave':
featurizer = deepchem.feat.WeaveFeaturizer()
elif featurizer == 'Raw':
featurizer = deepchem.feat.RawFeaturizer()
elif featurizer == 'AdjacencyConv':
featurizer = deepchem.feat.AdjacencyFingerprint(max_n_atoms=150,
max_valence=6)
elif featurizer == 'SelfDefine':
featurizer = deepchem.feat.UserDefinedFeaturizer(feature_field)
loader = deepchem.data.CSVLoader(tasks=m,
smiles_field="SMILES",
featurizer=featurizer)
dataset = loader.featurize(dataset_file,
data_dir=save_dir,
shard_size=8192)
# dataset = loader.featurize(dataset_file, shard_size=8192)
# Initialize transformers
if transformers:
transformers = [
deepchem.trans.NormalizationTransformer(transform_w=True,
dataset=dataset)
]
for transformer in transformers:
dataset = transformer.transform(dataset)
splitters = {
'index': deepchem.splits.IndexSplitter(),
'random': deepchem.splits.RandomSplitter(),
'random_stratified': deepchem.splits.RandomStratifiedSplitter(),
'scaffold': deepchem.splits.ScaffoldSplitter(),
'butina': deepchem.splits.ButinaSplitter(),
'task': deepchem.splits.TaskSplitter(),
'Harmonious_positive': Harmonious_positive(),
'OnePositiveSplit': OnePositiveSplit()
}
splitter = splitters[sep]
if sep == 'task':
fold_datasets = splitter.k_fold_split(dataset, K)
all_dataset = fold_datasets
elif sep == 'Harmonious_positive':
train_dataset, valid_dataset, test_dataset = splitter.train_valid_test_split(
dataset)
train_dataset = DiskDataset.from_numpy(train_dataset.X,
train_dataset.y,
train_dataset.w,
train_dataset.ids,
dataset.tasks,
data_dir=train)
valid_dataset = DiskDataset.from_numpy(valid_dataset.X,
valid_dataset.y,
valid_dataset.w,
valid_dataset.ids,
dataset.tasks,
data_dir=valid)
test_dataset = DiskDataset.from_numpy(test_dataset.X,
test_dataset.y,
test_dataset.w,
test_dataset.ids,
dataset.tasks,
data_dir=test)
all_dataset = (dataset, train_dataset, valid_dataset, test_dataset)
elif sep == 'Harmonious_positive' and K:
# train_dataset, valid_dataset, test_dataset = splitter.train_valid_test_split(
# dataset,
# frac_train=frac_train,
# frac_valid=0,
# frac_test=1- frac_train,
# )
# train_dataset = DiskDataset.from_numpy(train_dataset.X,train_dataset.y,train_dataset.w,train_dataset.ids,
# dataset.tasks,data_dir=train)
# train_dataset.reshard(8192)
# test_dataset = DiskDataset.from_numpy(test_dataset.X,test_dataset.y,test_dataset.w,test_dataset.ids,
# dataset.tasks,data_dir=test)
# test_dataset.reshard(8192)
# fold_dataset = splitter.k_fold_split(
# train_dataset, K, directories=[os.path.join(valid,str(i)) for i in range(K)],verbose=True)
fold_dataset = splitter.k_fold_split(
dataset,
K,
directories=[os.path.join(valid, str(i)) for i in range(K)],
verbose=True)
folds = []
for i in range(K):
print('merge fold dataset {}...'.format(i))
train_fold = DiskDataset.merge(
[fold_dataset[j] for j in range(K) if j != i],
merge_dir=os.path.join(valid, str(i), 'train_fold'))
test_fold = DiskDataset.merge([fold_dataset[i]],
merge_dir=os.path.join(
valid, str(i), 'valid_fold'))
folds.append([train_fold, test_fold])
all_dataset = (dataset, [], folds, [])
else:
train_dataset, valid_dataset, test_dataset = splitter.train_valid_test_split(
dataset,
train_dir=train,
valid_dir=valid,
test_dir=test,
frac_train=frac_train,
frac_valid=frac_valid,
frac_test=frac_test)
all_dataset = (dataset, train_dataset, valid_dataset, test_dataset)
# else:
# train_dataset, valid_dataset, test_dataset = splitter.train_valid_test_split(dataset,train_dir=train, valid_dir=valid, test_dir=test)
# all_dataset = (dataset,train_dataset, valid_dataset, test_dataset)
# if reload:
# deepchem.utils.save.save_dataset_to_disk(save_dir, train, valid, test,transformers)
return m, all_dataset, transformers
def k_fold_split(self, dataset, k, directories=None, **kwargs):
"""
Parameters
----------
dataset: Dataset
Dataset to do a k-fold split
k: int
number of folds
directories: list of str
list of length 2*k filepaths to save the result disk-datasets
kwargs
Returns
-------
list of length k tuples of (train, cv)
"""
"""
:param dataset:
:param k:
:param directories:
:param kwargs:
:return: list of length k tuples of (train, cv)
"""
log("Computing K-fold split", self.verbose)
if directories is None:
directories = [tempfile.mkdtemp() for _ in range(2 * k)]
else:
assert len(directories) == 2 * k
cv_datasets = []
train_ds_base = None
train_datasets = []
# rem_dataset is remaining portion of dataset
if isinstance(dataset, DiskDataset):
rem_dataset = dataset
else:
rem_dataset = DiskDataset.from_numpy(dataset.X, dataset.y, dataset.w,
dataset.ids)
for fold in range(k):
# Note starts as 1/k since fold starts at 0. Ends at 1 since fold goes up
# to k-1.
frac_fold = 1. / (k - fold)
train_dir, cv_dir = directories[2 * fold], directories[2 * fold + 1]
fold_inds, rem_inds, _ = self.split(
rem_dataset,
frac_train=frac_fold,
frac_valid=1 - frac_fold,
frac_test=0,
**kwargs)
cv_dataset = rem_dataset.select(fold_inds, select_dir=cv_dir)
cv_datasets.append(cv_dataset)
rem_dataset = rem_dataset.select(rem_inds)
train_ds_to_merge = filter(lambda x: x is not None,
[train_ds_base, rem_dataset])
train_ds_to_merge = filter(lambda x: len(x) > 0, train_ds_to_merge)
train_dataset = DiskDataset.merge(train_ds_to_merge, merge_dir=train_dir)
train_datasets.append(train_dataset)
update_train_base_merge = filter(lambda x: x is not None,
[train_ds_base, cv_dataset])
train_ds_base = DiskDataset.merge(update_train_base_merge)
return list(zip(train_datasets, cv_datasets))
def train_test_mtnn(train_task_csvs,
test_tasks_csvs,
tasks_nickname,
smiles_field,
y_field,
id_field,
tempdir,
num_epochs=40,
batch_size=128,
learning_rate=0.001,
graph_conv_sizes=(128, 128),
dense_size=256,
gpu=None):
"""
Trains a multitask GCNN using the training sets in train_tasks_csvs and validates it using the test sets in
test_tasks_csvs. Saves the trained model and the predictions under a folder named "train_test". Prints performance
metrics (R2 and Spearman rho) after every epoch.
NB: each task in the model should have a corresponding training and test files, named similarly (ex:task1_train.csv,
task1_test.csv).
:param train_task_csvs: list of csv files containing the training tasks
:param test_tasks_csvs: list of csv files containing the test tasks
:param tasks_nickname: how the model will be named
:param smiles_field: in the csvs, name of the column containing the smiles string of the cpds
:param y_field: in the csv, name of the column containing the activity data
:param id_field: in the csv, name of the column containing the molids
:param tempdir: where to store the temporary files for the DiskDatasets (will be deleted later on)
:param num_epochs: how many epochs to train for
:param batch_size: number of molecules per minibatch
:param learning_rate: learning rate
:param graph_conv_sizes: tuple with output dimension for every GC layer
:param dense_size: nb of neurons in the last dense layer
:param gpu: GPU to use for training (if None, only CPU will be used)
:return: None
"""
ensure_dir(tempdir)
tasks, training_dset = load_training_data(train_task_csvs,
smiles_field=smiles_field,
y_field=y_field,
id_field=id_field,
tempdir=op.join(
tempdir, 'train'),
cv=False)
tasks, test_dset = load_training_data(test_tasks_csvs,
smiles_field=smiles_field,
y_field=y_field,
id_field=id_field,
tempdir=op.join(tempdir, 'test'),
cv=False)
# Take care of outdir
outdir = get_multitask_traintest_outdir(tasks_nickname)
ensure_dir(outdir)
# Have we already run that experiment?
if op.exists(op.join(outdir, 'DONE.txt')):
print('Model already trained and validated.')
else:
print('Training and validating multitask graph convolution model')
# Merge to reduce the number of shards (very necessary to avoid weird problems of non-random minibatches)
disk_dir_to_delete = tempfile.mkdtemp(prefix=tempdir + '/')
training_dset = DiskDataset.merge([training_dset],
merge_dir=disk_dir_to_delete)
# Transformation (z-scaling)
zscaling_dir_train = op.join(tempdir, 'zscaling', 'train')
ensure_dir(zscaling_dir_train)
zscaling_dir_test = op.join(tempdir, 'zscaling', 'test')
ensure_dir(zscaling_dir_test)
transfo_dir_to_delete_1 = tempfile.mkdtemp(prefix=zscaling_dir_train +
'/')
transfo_dir_to_delete_2 = tempfile.mkdtemp(prefix=zscaling_dir_test +
'/')
transformer = NormalizationTransformer(transform_y=True,
dataset=training_dset)
scaled_train = transformer.transform(training_dset,
outdir=transfo_dir_to_delete_1)
scaled_val = transformer.transform(test_dset,
outdir=transfo_dir_to_delete_2)
# Train the model
scaled_train_y, yhattrain, scaled_train_w, scaled_test_y, yhattest, scaled_test_w = \
train_and_validate_mtnn(scaled_train, n_tasks=len(tasks), outdir=outdir, graph_conv_sizes=graph_conv_sizes,
dense_size=dense_size, batch_size=batch_size, learning_rate=learning_rate,
num_epochs=num_epochs, pickle_file_name=tasks_nickname + '.pkl', test=scaled_val,
transformer=transformer, test_unscaled=test_dset, gpu=gpu)
# compute metrics
scaled_results_test = evaluate_multitask_gc(scaled_test_y, yhattest,
scaled_test_w)
for k, vals in scaled_results_test.items():
print(k)
print(vals)
# let's reverse the transformation from the predictions
yhattest_untransf = undo_transforms(yhattest, [transformer])
unscaled_results_test = evaluate_multitask_gc(test_dset.y,
yhattest_untransf,
test_dset.w)
for k, vals in unscaled_results_test.items():
print(k)
print(vals)
# hopefully the results are very similar
# Remove transfo dir
shutil.rmtree(transfo_dir_to_delete_1)
shutil.rmtree(transfo_dir_to_delete_2)
# Get rid of the temporary directory structure
shutil.rmtree(tempdir)
print('Dataset folders removed!')
def cross_validate_mtnn(task_csvs,
tasks_nickname,
smiles_field,
split_field,
y_field,
id_field,
tempdir,
num_epochs,
batch_size=128,
learning_rate=0.001,
graph_conv_sizes=(128, 128),
dense_size=256,
gpu=None):
"""
Cross-validates a multitask GCNN using the training sets in train_tasks_csvs. Saves the trained models and the
predictions under folders named "fold_i". Prints performance metrics (R2 and Spearman rho) after every epoch.
NB: each task in the model should have a corresponding training file. A columns with fold assignment should be
provided for the cross-validation.
:param task_csvs: list of csv files containing the training tasks
:param tasks_nickname: how the model will be named
:param smiles_field: in the csvs, name of the column containing the smiles string of the cpds
:param split_field: in the csvs, name of the column containing the fold assignment for the cross-validation
:param y_field: in the csvs, name of the column containing the activity data
:param id_field: in the csv, name of the column containing the molids
:param tempdir: where to store the temporary files for the DiskDatasets (will be deleted later on)
:param num_epochs: how many epochs to train for
:param batch_size: number of molecules per minibatch
:param learning_rate: learning rate
:param graph_conv_sizes: tuple with output dimension for every GC layer
:param dense_size: nb of neurons in the last dense layer
:param gpu: GPU to use for training (if None, only CPU will be used)
:return: A pandas dataframe with performance metrics for every fold
"""
ensure_dir(tempdir)
tasks, folds, fold_dirs = load_training_data(task_csvs,
smiles_field=smiles_field,
split_field=split_field,
y_field=y_field,
id_field=id_field,
tempdir=tempdir,
cv=True)
fold_results = defaultdict(list)
for i, fold in enumerate(folds):
# Take care of outdir
outdir = get_multitask_cv_outdir(tasks_nickname, i)
ensure_dir(outdir)
# Have we already run that fold?
if op.exists(op.join(outdir, 'DONE.txt')):
print('Fold %i already computed.' % i)
else:
print('Running graph convolution model for fold %i' % i)
val = fold
disk_dir_to_delete = tempfile.mkdtemp(prefix=tempdir + '/')
train = DiskDataset.merge(folds[0:i] + folds[i + 1:],
merge_dir=disk_dir_to_delete)
# Transformation (z-scaling)
zscaling_dir_train = op.join(tempdir, 'zscaling', 'train')
ensure_dir(zscaling_dir_train)
zscaling_dir_test = op.join(tempdir, 'zscaling', 'test')
ensure_dir(zscaling_dir_test)
transfo_dir_to_delete_1 = tempfile.mkdtemp(
prefix=zscaling_dir_train + '/')
transfo_dir_to_delete_2 = tempfile.mkdtemp(
prefix=zscaling_dir_test + '/')
transformer = NormalizationTransformer(transform_y=True,
dataset=train)
scaled_train = transformer.transform(
train, outdir=transfo_dir_to_delete_1)
scaled_val = transformer.transform(val,
outdir=transfo_dir_to_delete_2)
train_y, yhattrain, train_w, test_y, yhattest, test_w = \
train_and_validate_mtnn(scaled_train, len(tasks), outdir=outdir, graph_conv_sizes=graph_conv_sizes,
dense_size=dense_size, batch_size=batch_size, learning_rate=learning_rate,
num_epochs=num_epochs, pickle_file_name=tasks_nickname + '_fold_%i.pkl' % i,
test=scaled_val, test_unscaled=val, transformer=transformer, fold=i, gpu=gpu)
# compute metrics
train_results = evaluate_multitask_gc(train_y, yhattrain, train_w)
test_results = evaluate_multitask_gc(test_y, yhattest, test_w)
# Populate the results dictionary
for j, t in enumerate(tasks):
fold_results['fold'].append(i)
fold_results['task'].append(t)
fold_results['train'].append(True)
fold_results['r2'].append(train_results[j][0])
fold_results['mse'].append(train_results[j][1])
fold_results['mae'].append(train_results[j][2])
fold_results['varex'].append(train_results[j][3])
fold_results['spearman'].append(train_results[j][4])
fold_results['fold'].append(i)
fold_results['task'].append(t)
fold_results['train'].append(False)
fold_results['r2'].append(test_results[j][0])
fold_results['mse'].append(test_results[j][1])
fold_results['mae'].append(test_results[j][2])
fold_results['varex'].append(test_results[j][3])
fold_results['spearman'].append(test_results[j][4])
# Clean the tempdirs
shutil.rmtree(disk_dir_to_delete)
shutil.rmtree(transfo_dir_to_delete_1)
shutil.rmtree(transfo_dir_to_delete_2)
print('folder removed!')
# Get rid of the foldirs
for foldir in fold_dirs:
shutil.rmtree(foldir)
shutil.rmtree(tempdir)
print('fold dataset folders removed!')
return pd.DataFrame.from_dict(fold_results)
def k_fold_split(self, dataset, k, directories=None, **kwargs):
"""
Parameters
----------
dataset: Dataset
Dataset to do a k-fold split
k: int
number of folds
directories: list of str
list of length 2*k filepaths to save the result disk-datasets
kwargs
Returns
-------
list of length k tuples of (train, cv)
"""
"""
:param dataset:
:param k:
:param directories:
:param kwargs:
:return: list of length k tuples of (train, cv)
"""
log("Computing K-fold split", self.verbose)
if directories is None:
directories = [tempfile.mkdtemp() for _ in range(2 * k)]
else:
assert len(directories) == 2 * k
cv_datasets = []
train_ds_base = None
train_datasets = []
# rem_dataset is remaining portion of dataset
rem_dataset = dataset
for fold in range(k):
# Note starts as 1/k since fold starts at 0. Ends at 1 since fold goes up
# to k-1.
frac_fold = 1. / (k - fold)
train_dir, cv_dir = directories[2 * fold], directories[2 * fold +
1]
fold_inds, rem_inds, _ = self.split(rem_dataset,
frac_train=frac_fold,
frac_valid=1 - frac_fold,
frac_test=0)
cv_dataset = rem_dataset.select(fold_inds)
cv_datasets.append(cv_dataset)
rem_dataset = rem_dataset.select(rem_inds)
train_ds_to_merge = filter(lambda x: x is not None,
[train_ds_base, rem_dataset])
train_ds_to_merge = filter(lambda x: len(x) > 0, train_ds_to_merge)
train_dataset = DiskDataset.merge(train_ds_to_merge,
merge_dir=train_dir)
train_datasets.append(train_dataset)
update_train_base_merge = filter(lambda x: x is not None,
[train_ds_base, cv_dataset])
train_ds_base = DiskDataset.merge(update_train_base_merge,
merge_dir=cv_dir)
return list(zip(train_datasets, cv_datasets))