def write_feature_matrix(): seqs_h1n1, seqs_h3n2, seqs_h5n1, seqs_h9n2, mixed_seqs = load_aligned_seqs() dh1n1, dh3n2, dh5n1, dh9n2, dmixed = mix_subtypes_dist() h1n1_mutMatrix = get_feature_matrix(seqs_h1n1, dh1n1, featureMatrix_path + "h1n1.feats") h3n2_mutMatrix = get_feature_matrix(seqs_h3n2, dh3n2, featureMatrix_path + "h3n2.feats") h5n1_mutMatrix = get_feature_matrix(seqs_h5n1, dh5n1, featureMatrix_path + "h5n1.feats") h9n2_mutMatrix = get_feature_matrix(seqs_h9n2, dh9n2, featureMatrix_path + "h9n2.feats") mixed_mutMatrix = get_feature_matrix(mixed_seqs, dmixed, featureMatrix_path + "mixed.feats")
def transfer_learning(test_data):
# Train on any three datasets and test on the rest
# e.g. train on H1N1+H3N2+H5N1; test on H9N2
print test_data
dh1n1, dh3n2, dh5n1, dh9n2, dmixed = mix_subtypes_dist()
h1n1_mutMatrix_file = mutMatrix_path + "h1n1.mut"
h3n2_mutMatrix_file = mutMatrix_path + "h3n2.mut"
h5n1_mutMatrix_file = mutMatrix_path + "h5n1.mut"
h9n2_mutMatrix_file = mutMatrix_path + "h9n2.mut"
X_h1n1, Y_h1n1, XName_h1n1 = load_mut_dataset(h1n1_mutMatrix_file, dh1n1)
X_h3n2, Y_h3n2, XName_h3n2 = load_mut_dataset(h3n2_mutMatrix_file, dh3n2)
X_h5n1, Y_h5n1, XName_h5n1 = load_mut_dataset(h5n1_mutMatrix_file, dh5n1)
X_h9n2, Y_h9n2, XName_h9n2 = load_mut_dataset(h9n2_mutMatrix_file, dh9n2)
dataset = {}
dataset["h1n1"] = (X_h1n1, Y_h1n1, XName_h1n1)
dataset["h3n2"] = (X_h3n2, Y_h3n2, XName_h3n2)
dataset["h5n1"] = (X_h5n1, Y_h5n1, XName_h5n1)
dataset["h9n2"] = (X_h9n2, Y_h9n2, XName_h9n2)
tmp = ["h1n1", "h3n2", "h5n1", "h9n2"]
tmp.remove(test_data)
(X_test, Y_test, XName_test) = dataset[test_data]
X_test = np.array(X_test)
Y_test = np.array(Y_test)
XName_test = np.array(XName_test)
labels_test = create_binary_labels(Y_test, 4)
(X1, Y1, XName1) = dataset[tmp[0]]
(X2, Y2, XName2) = dataset[tmp[1]]
(X3, Y3, XName3) = dataset[tmp[2]]
X_train = np.concatenate((X1, X2, X3), axis=0)
Y_train = np.concatenate((Y1, Y2, Y3), axis=0)
XName_train = np.concatenate((XName1, XName2, XName3), axis=0)
labels_train = create_binary_labels(Y_train, 4)
for i in range(10):
outfile = "../result/bioinfo2008Liao/tl_" + test_data + "_" + str(
i) + ".gm5"
X_train, Y_train, labels_train, XName_train = shuffle_dataset_with_labels(
X_train, Y_train, labels_train, XName_train)
X_test, Y_test, labels_test, XName_test = shuffle_dataset_with_labels(
X_test, Y_test, labels_test, XName_test)
result = multiReg_GM5(outfile, X_train, Y_train, labels_train,
XName_train, X_test, Y_test, labels_test,
XName_test)
def writeMutMatrix():
seqs_h1n1, seqs_h3n2, seqs_h5n1, seqs_h9n2, mixed_seqs = load_aligned_seqs(
)
dh1n1, dh3n2, dh5n1, dh9n2, dmixed = mix_subtypes_dist()
# mutMatrix_path = "../data/bioinfo2008Liao_mutMatrix/"
h1n1_mutMatrix = get_mutMatrix(seqs_h1n1, dh1n1,
mutMatrix_path + "h1n1.mut")
h3n2_mutMatrix = get_mutMatrix(seqs_h3n2, dh3n2,
mutMatrix_path + "h3n2.mut")
h5n1_mutMatrix = get_mutMatrix(seqs_h5n1, dh5n1,
mutMatrix_path + "h5n1.mut")
h9n2_mutMatrix = get_mutMatrix(seqs_h9n2, dh9n2,
mutMatrix_path + "h9n2.mut")
mixed_mutMatrix = get_mutMatrix(mixed_seqs, dmixed,
mutMatrix_path + "mixed.mut")
def transfer_learning(test_data):
# print "transfer learning test on: " + test_data
dh1n1, dh3n2, dh5n1, dh9n2, dmixed = mix_subtypes_dist()
h1n1_fm_file = featureMatrix_path + "h1n1.feats"
h3n2_fm_file = featureMatrix_path + "h3n2.feats"
h5n1_fm_file = featureMatrix_path + "h5n1.feats"
h9n2_fm_file = featureMatrix_path + "h9n2.feats"
X_h1n1, Y_h1n1, XName_h1n1 = load_features_dataset(h1n1_fm_file, dh1n1)
X_h3n2, Y_h3n2, XName_h3n2 = load_features_dataset(h3n2_fm_file, dh3n2)
X_h5n1, Y_h5n1, XName_h5n1 = load_features_dataset(h5n1_fm_file, dh5n1)
X_h9n2, Y_h9n2, XName_h9n2 = load_features_dataset(h9n2_fm_file, dh9n2)
dataset = {}
dataset["h1n1"] = (X_h1n1, Y_h1n1, XName_h1n1)
dataset["h3n2"] = (X_h3n2, Y_h3n2, XName_h3n2)
dataset["h5n1"] = (X_h5n1, Y_h5n1, XName_h5n1)
dataset["h9n2"] = (X_h9n2, Y_h9n2, XName_h9n2)
tmp = ["h1n1", "h3n2", "h5n1", "h9n2"]
tmp.remove(test_data)
(X_test, Y_test, XName_test) = dataset[test_data]
X_test = np.array(X_test)
Y_test = np.array(Y_test)
XName_test = np.array(XName_test)
labels_test = create_binary_labels(Y_test, 4)
(X1, Y1, XName1) = dataset[tmp[0]]
(X2, Y2, XName2) = dataset[tmp[1]]
(X3, Y3, XName3) = dataset[tmp[2]]
X_train = np.concatenate((X1, X2, X3), axis=0)
Y_train = np.concatenate((Y1, Y2, Y3), axis=0)
XName_train = np.concatenate((XName1, XName2, XName3), axis=0)
labels_train = create_binary_labels(Y_train, 4)
for i in range(10):
outfile = "../result/srep2017Adapted/tl_" + test_data + "_" + str(
i) + ".rf_srep"
X_train, Y_train, labels_train, XName_train = shuffle_dataset_with_labels(
X_train, Y_train, labels_train, XName_train)
X_test, Y_test, labels_test, XName_test = shuffle_dataset_with_labels(
X_test, Y_test, labels_test, XName_test)
result = rf_model(outfile, X_train, Y_train, labels_train, XName_train,
X_test, Y_test, labels_test, XName_test)
def transfer_learning_2():
# Train on H1N1+H3N2; Test on H5N1+H9N2
dh1n1, dh3n2, dh5n1, dh9n2, dmixed = mix_subtypes_dist()
h1n1_fm_file = featureMatrix_path + "h1n1.feats"
h3n2_fm_file = featureMatrix_path + "h3n2.feats"
h5n1_fm_file = featureMatrix_path + "h5n1.feats"
h9n2_fm_file = featureMatrix_path + "h9n2.feats"
X_h1n1, Y_h1n1, XName_h1n1 = load_features_dataset(h1n1_fm_file, dh1n1)
X_h3n2, Y_h3n2, XName_h3n2 = load_features_dataset(h3n2_fm_file, dh3n2)
X_h5n1, Y_h5n1, XName_h5n1 = load_features_dataset(h5n1_fm_file, dh5n1)
X_h9n2, Y_h9n2, XName_h9n2 = load_features_dataset(h9n2_fm_file, dh9n2)
dataset = {}
dataset["h1n1"] = (X_h1n1, Y_h1n1, XName_h1n1)
dataset["h3n2"] = (X_h3n2, Y_h3n2, XName_h3n2)
dataset["h5n1"] = (X_h5n1, Y_h5n1, XName_h5n1)
dataset["h9n2"] = (X_h9n2, Y_h9n2, XName_h9n2)
Xtrain2 = np.concatenate((X_h1n1, X_h3n2), axis=0)
Ytrain2 = np.concatenate((Y_h1n1, Y_h3n2), axis=0)
XNameTrain2 = np.concatenate((XName_h1n1, XName_h3n2), axis=0)
Xtest2 = np.concatenate((X_h5n1, X_h9n2), axis=0)
Ytest2 = np.concatenate((Y_h5n1, Y_h9n2), axis=0)
XNameTest2 = np.concatenate((XName_h5n1, XName_h9n2), axis=0)
labels_train2 = create_binary_labels(Ytrain2, 4)
labels_test2 = create_binary_labels(Ytest2, 4)
for i in range(10):
outfile = "../result/srep2017Adapted/tl2_" + str(i) + ".rf_srep"
X_train, Y_train, labels_train, XName_train = shuffle_dataset_with_labels(
Xtrain2, Ytrain2, labels_train2, XNameTrain2)
X_test, Y_test, labels_test, XName_test = shuffle_dataset_with_labels(
Xtest2, Ytest2, labels_test2, XNameTest2)
result = rf_model(outfile, X_train, Y_train, labels_train, XName_train,
X_test, Y_test, labels_test, XName_test)
def transfer_learning_2():
dh1n1, dh3n2, dh5n1, dh9n2, dmixed = mix_subtypes_dist()
h1n1_mutMatrix_file = mutMatrix_path + "h1n1.mut"
h3n2_mutMatrix_file = mutMatrix_path + "h3n2.mut"
h5n1_mutMatrix_file = mutMatrix_path + "h5n1.mut"
h9n2_mutMatrix_file = mutMatrix_path + "h9n2.mut"
X_h1n1, Y_h1n1, XName_h1n1 = load_mut_dataset(h1n1_mutMatrix_file, dh1n1)
X_h3n2, Y_h3n2, XName_h3n2 = load_mut_dataset(h3n2_mutMatrix_file, dh3n2)
X_h5n1, Y_h5n1, XName_h5n1 = load_mut_dataset(h5n1_mutMatrix_file, dh5n1)
X_h9n2, Y_h9n2, XName_h9n2 = load_mut_dataset(h9n2_mutMatrix_file, dh9n2)
dataset = {}
dataset["h1n1"] = (X_h1n1, Y_h1n1, XName_h1n1)
dataset["h3n2"] = (X_h3n2, Y_h3n2, XName_h3n2)
dataset["h5n1"] = (X_h5n1, Y_h5n1, XName_h5n1)
dataset["h9n2"] = (X_h9n2, Y_h9n2, XName_h9n2)
# Train on H1N1+H3N2; Test on H5N1+H9N2
Xtrain2 = np.concatenate((X_h1n1, X_h3n2), axis=0)
Ytrain2 = np.concatenate((Y_h1n1, Y_h3n2), axis=0)
XNameTrain2 = np.concatenate((XName_h1n1, XName_h3n2), axis=0)
Xtest2 = np.concatenate((X_h5n1, X_h9n2), axis=0)
Ytest2 = np.concatenate((Y_h5n1, Y_h9n2), axis=0)
XNameTest2 = np.concatenate((XName_h5n1, XName_h9n2), axis=0)
labels_train2 = create_binary_labels(Ytrain2, 4)
labels_test2 = create_binary_labels(Ytest2, 4)
for i in range(10):
outfile = "../result/bioinfo2008Liao/tl2_" + str(i) + ".gm5"
X_train, Y_train, labels_train, XName_train = shuffle_dataset_with_labels(
Xtrain2, Ytrain2, labels_train2, XNameTrain2)
X_test, Y_test, labels_test, XName_test = shuffle_dataset_with_labels(
Xtest2, Ytest2, labels_test2, XNameTest2)
result = multiReg_GM5(outfile, X_train, Y_train, labels_train,
XName_train, X_test, Y_test, labels_test,
XName_test)
line = prefix + ", " + str(avg_accu1) + ", " + str(avg_accu2) + ", "
line = line + str(avg_prec1) + ", " + str(avg_prec2) + ", "
line = line + str(avg_rec1) + ", " + str(avg_rec2) + ", "
line = line + str(avg_f1_1) + ", " + str(avg_f1_2) + "\n"
result_csv.write(line)
result_csv.close()
if __name__ == '__main__':
seqs_H1N1, seqs_H3N2, seqs_H5N1, seqs_H9N2, mixed_seqs = load_aligned_seqs(
)
dh1n1, dh3n2, dh5n1, dh9n2, dmixed = mix_subtypes_dist()
train_rfsrep("h1n1", dh1n1)
train_rfsrep("h3n2", dh3n2)
train_rfsrep("h5n1", dh5n1)
train_rfsrep("h9n2", dh9n2)
train_rfsrep("mixed", dmixed)
transfer_learning_2()
transfer_learning("h1n1")
transfer_learning("h3n2")
transfer_learning("h5n1")
transfer_learning("h9n2")
writeResults_csv()
