def hide_non_minimal_complexes(pdb_object):
scores = rosetta.parse_scores(pdb_object.minimized.scores.read())
hidden_folder = pdb_object.minimized.hidden_complexes.path
hidden_folder.delete()
storage.make_directory(hidden_folder)
for number in list(scores.keys())[1:]:
complex_path = pdb_object.minimized.complex.pdb[number].path
storage.move(complex_path, hidden_folder, no_fail=True)
def hide_non_minimal_complexes(pdb_object):
"""Generate constraint file for Rosetta minimization.
Constraints are added to maintain contact ions close to their original positions.
pdb_object : PDBObject
PDB structure to be handled
"""
scores = rosetta.parse_scores(pdb_object.minimized.scores.read())
hidden_folder = pdb_object.minimized.hidden_complexes.path
storage.make_directory(hidden_folder)
for number in list(scores.keys())[1:]:
complex_path = pdb_object.minimized.complex.pdb[number].path
print("Hiding ", complex_path)
storage.move(complex_path, hidden_folder, no_fail=True)
def generate_tfrecords(dataset_object):
"""Generate TFRecords from a dataset's combined images.
dataset_object : DatasetObject
Dataset of the images.
"""
files = dataset_object.images
chunks = chunk_by_size(files)
storage.clear_directory(dataset_object.tfrecords, no_fail=True)
storage.make_directory(dataset_object.tfrecords, no_fail=True)
lines = dataset_object.labels.read().splitlines()
lines = [line.split(" ") for line in lines]
pdb_labels = dict([(line[0], line[1:]) for line in lines])
for i, chunk in enumerate(chunks):
write_tfrecords(chunk, dataset_object, i, pdb_labels)
def combine_maps(pdb_object):
"""Postprocessing step for combining the feature map images into one image
Parameters
----------
pdb_object : PDBObject
PDB object which images will be combined
"""
features = [
pdb_object.image.htmd.read(),
pdb_object.image.electronegativity.read(),
pdb_object.image.rosetta.read()
]
grid = np.concatenate(features, axis=-1)
storage.make_directory(pdb_object.image.combined.path.parent)
pdb_object.image.combined.write(grid)
def __init__(self, model_object, dataset_object, channels, action, seed):
self.dataset_object = dataset_object
if not self.results.path.exists():
self.results.write("")
self.model_object = model_object
self.channels = channels
input_fn = input.load_fn(dataset_object, channels, settings.rotate,
action)
self.iterator = input_fn().make_initializable_iterator()
model_fn = model_object.load_fn()
self.id, self.X, self.y = self.iterator.get_next()
self.shape = tf.shape(self.y)[0]
model = model_fn(self.X, self.y, action, settings.rotate)
self.op = model.train_op
self.loss = model.loss
self.predictions = model.predictions
checkpoint_folder = self.dataset_object.model(self.model_object,
self.channels, seed)
storage.make_directory(checkpoint_folder)
self.save_path = checkpoint_folder / "model.ckpt"
self.saver = tf.train.Saver()