dragonn.models module

class dragonn.models.DecisionTree[source]

Bases: dragonn.models.Model

predict(X)[source]
train(X, y, validation_data=None)[source]
class dragonn.models.Model(**hyperparameters)[source]

Bases: object

predict(X)[source]
score(X, y, metric)[source]
test(X, y)[source]
train(X, y, validation_data)[source]
class dragonn.models.MotifScoreRNN(input_shape, gru_size=10, tdd_size=4)[source]

Bases: dragonn.models.Model

predict(X)[source]
train(X, y, validation_data)[source]
class dragonn.models.RandomForest[source]

Bases: dragonn.models.DecisionTree

class dragonn.models.SVC[source]

Bases: dragonn.models.Model

predict(X)[source]
train(X, y, validation_data=None)[source]
class dragonn.models.SequenceDNN(seq_length, use_deep_CNN=False, use_RNN=False, num_tasks=1, num_filters=15, conv_width=15, num_filters_2=15, conv_width_2=15, num_filters_3=15, conv_width_3=15, pool_width=35, L1=0, dropout=0.0, GRU_size=35, TDD_size=15, verbose=1)[source]

Bases: dragonn.models.Model

Sequence DNN models.

seq_length : int
length of input sequence.
use_deep_CNN : bool, optional
uses 3 layered CNN if True, 1 layered CNN if False. Default: False.
num_tasks : int,
number of tasks. Default: 1.
num_filters : int
number of 1st layer convolutional filters. Default: 15.
conv_width : int
width of 1st layer convolutional filters. Default: 15.
pool_width : int
width of max pooling. Default: 35.
num_filters_2 : int
number of 2nd layer convolutional filters. Default: 15.
conv_width_2 : int
width of 2nd layer convolutional filters. Default: 15.
num_filters_3 : int
number of 3rd layer convolutional filters. Default: 15.
conv_width_3 : int
width of 3rd layer convolutional filters. Default: 15.
L1 : float
strength of L1 penalty.
dropout : float
dropout probability in every convolutional layer. Default: 0.
num_tasks : int
Number of prediction tasks or labels. Default: 1.
verbose: int
Verbosity level during training. Valida values: 0, 1, 2.

Compiled DNN model.

class LossHistory(X_train, y_train, validation_data, sequence_DNN)[source]

Bases: keras.callbacks.Callback

on_epoch_end(epoch, logs={})[source]
class SequenceDNN.PrintMetrics(validation_data, sequence_DNN)[source]

Bases: keras.callbacks.Callback

on_epoch_end(epoch, logs={})[source]
SequenceDNN.deeplift(X, batch_size=200)[source]

Returns (num_task, num_samples, input_shape) deeplift score array.

SequenceDNN.get_sequence_filters()[source]

Returns list with sequence filter 2darrays.

SequenceDNN.in_silico_mutagenesis(X)[source]
SequenceDNN.predict(X)[source]
SequenceDNN.train(X, y, validation_data)[source]
class dragonn.models.gkmSVM(prefix='./gkmSVM', word_length=11, mismatches=3, C=1)[source]

Bases: dragonn.models.Model

static encode_sequence_into_fasta_file(sequence_iterator, ofname)[source]

writes sequences into fasta file

model_file
predict(X)[source]
train(X, y, validation_data=None)[source]

Trains gkm-svm, saves model file.