def test_encoder_index_lists():
encoder = Encoder()
S_idx = encoder.index_dict["S"]
A_idx = encoder.index_dict["A"]
index_lists = encoder.encode_index_lists(["SSS", "AAA", "SAS"])
eq_(index_lists,
[[S_idx, S_idx, S_idx], [A_idx, A_idx, A_idx], [S_idx, A_idx, S_idx]])
def test_encoder_index_array():
encoder = Encoder()
S_idx = encoder.index_dict["S"]
A_idx = encoder.index_dict["A"]
assert S_idx > 0
assert A_idx > 0
X = encoder.encode_index_array(["SSS", "AAA", "SASA"],
max_peptide_length=4)
expected = np.array([[S_idx, S_idx, S_idx, 0], [A_idx, A_idx, A_idx, 0],
[S_idx, A_idx, S_idx, A_idx]])
assert (X == expected).all()
def test_model_with_fixed_length_context():
model = make_variable_length_model_with_fixed_length_context(
n_upstream=1, n_downstream=1, max_peptide_length=3)
encoder = Encoder()
X_peptide = encoder.encode_index_array(["SYF", "QQ", "C", "GLL"],
max_peptide_length=3)
X_upstream = encoder.encode_index_array(["Q", "A", "L", "I"])
X_downstream = encoder.encode_index_array(["S"] * 4)
Y = np.array([True, False, True, False])
input_dict = {
"upstream": X_upstream,
"downstream": X_downstream,
"peptide": X_peptide
}
model.fit(input_dict, Y, epochs=20)
Y_pred = model.predict(input_dict)
assert (Y == (Y_pred[:, 0] > 0.5)).all(), (Y, Y_pred)
def test_encoder_FOFE_bidirectional():
# turn off the gap character '-' used for ends of shorter sequences
encoder = Encoder(variable_length_sequences=False)
x = encoder.encode_FOFE(["AAA", "SSS", "SASA"], bidirectional=True)
eq_(x.shape, (3, 40))
def test_encoder_prepare_sequences_stop_token():
encoder = Encoder(add_stop_tokens=True)
eq_(encoder.prepare_sequences(["SISI"], 5), ["SISI$-"])
def test_encoder_prepare_sequences_padding():
encoder = Encoder()
eq_(encoder.prepare_sequences(["SISI"], 5), ["SISI-"])
def test_encoder_onehot_with_positional_features():
encoder = Encoder(variable_length_sequences=False,
add_normalized_position=True,
add_normalized_centrality=True)
x = encoder.encode_onehot(["AAA", "SSS", "EEE"])
eq_(x.shape, (3, 3, 22))
def test_encoder_onehot():
encoder = Encoder(variable_length_sequences=False)
x = encoder.encode_onehot(["AAA", "SSS", "EEE"])
eq_(x.shape, (3, 3, 20))