def _compute_score(model, mirna_transcript, candidate_site, label_encoder,
one_hot_encoder):
"""
predict the score of the duplex passed as argument
:param model: the trained neural network model
:param label_encoder: the label encoder to use
:param one_hot_encoder: one_hot encoder object to use
:param mirna_transcript: the mirna transcript sequence
:param candidate_site: the candidate mbs sequence
:return: the score according to the model
"""
# reshape data
nn_input = np.concatenate(
(vec.one_hot_encode_sequence(mirna_transcript, label_encoder,
one_hot_encoder),
vec.one_hot_encode_sequence(candidate_site,
label_encoder,
one_hot_encoder,
mirna=False)))
nn_input = np.expand_dims(nn_input, axis=0)
# predict output
score = model.predict(nn_input)
return score
def test_one_hot_encode_sequence_site_len():
# check site
enc0, enc1 = init_encoders()
assert len(one_hot_encode_sequence('A' * 40, enc0, enc1,
mirna=False)) == 160
with pytest.raises(SystemExit):
one_hot_encode_sequence('A' * 41, enc0, enc1, mirna=False)
def test_one_hot_encode_sequence_mirna_len():
# check mirnas
enc0, enc1 = init_encoders()
assert len(one_hot_encode_sequence('', enc0, enc1)) == 120
assert len(one_hot_encode_sequence('U' * 25, enc0, enc1)) == 120
with pytest.raises(SystemExit):
one_hot_encode_sequence('U' * 31, enc0, enc1)
def test_one_hot_encode_sequence_site_wellformed():
# must not raise exception
enc0, enc1 = init_encoders()
raised = False
try:
one_hot_encode_sequence('A' * 40, enc0, enc1, mirna=False)
except SystemExit:
raised = True
assert raised == False
def test_one_hot_encode_sequence_site_badformed():
enc0, enc1 = init_encoders()
with pytest.raises(SystemExit):
one_hot_encode_sequence('H' * 40, enc0, enc1, mirna=False)
def test_one_hot_encode_sequence_mirna_badformed():
enc0, enc1 = init_encoders()
with pytest.raises(SystemExit):
one_hot_encode_sequence('H' * 20, enc0, enc1)