# loop over different models
for model in models:
print('model: ' + model)
model_path = helper.make_directory(sstype_path, model)
# loop over different RNA binding proteins
results = []
for rbp_index, experiment in enumerate(experiments):
print('Analyzing: ' + experiment)
tf.reset_default_graph() # reset any tensorflow graphs
np.random.seed(247) # for reproducibilitjy
tf.set_random_seed(247) # for reproducibility
# load rbp dataset
train, valid, test = helper.load_dataset_hdf5(data_path,
ss_type=ss_type,
rbp_index=rbp_index)
# process rbp dataset
train, valid, test = helper.process_data(train,
valid,
test,
method=normalize_method)
# get shapes
input_shape = list(train['inputs'].shape)
input_shape[0] = None
output_shape = train['targets'].shape
# load model
genome_model = helper.import_model(model)
# directory to save parameters of saliency
classifier_path = helper.make_directory(results_path, normalize_method+'_'+ss_type)
classifier_path = helper.make_directory(classifier_path, 'ensemble')
# path to pre-trained model parameters
best_path = os.path.join(trained_path, 'mixed_'+normalize_method+'_'+ss_type)
# loop over different RNA binding proteins
for rbp_name in rbp_names:
print('Analyzing: '+ rbp_name)
tf.reset_default_graph() # reset any tensorflow graphs
np.random.seed(247) # for reproducibilitjy
tf.set_random_seed(247) # for reproducibility
# load rbp dataset
train, valid, test = helper.load_dataset_hdf5(data_path, dataset_name=rbp_name, ss_type=ss_type)
# process rbp dataset
train, valid, test = helper.process_data(train, valid, test, method=normalize_method)
# get shapes
input_shape = list(train['inputs'].shape)
input_shape[0] = None
output_shape = train['targets'].shape
# get ensemble model predictions
ensemble_predictions, predictions = helper.ensemble_predictions(train, rbp_name, models,
input_shape, output_shape, best_path, use_scope=True)
max_index = np.argsort(ensemble_predictions)[::-1]
# directories to store classifier model and motifs
dataset_file_path = os.path.join(dataset_path,
experiment_name + '_' + str(window) + '.h5')
# set results path
results_path = helper.make_directory('../../results', 'encode_eclip')
# model results path
model_path = helper.make_directory(results_path, model_name)
# path to save model
experiment_name = rbp_name + '_' + cell_name
model_save_path = os.path.join(model_path, experiment_name)
#-----------------------------------------------------------------------------------------
# load dataset
train, valid, test = helper.load_dataset_hdf5(dataset_file_path,
ss_type=ss_type)
# get shapes of data
input_shape = list(train['inputs'].shape)
input_shape[0] = None
output_shape = [None, train['targets'].shape[1]]
# build model
genome_model = helper.import_model(model_name)
model_layers, optimization = genome_model(input_shape, output_shape)
# build neural network class
nnmodel = nn.NeuralNet(seed=247)
nnmodel.build_layers(model_layers, optimization)
nnmodel.inspect_layers()