def demo():
label_name_list = 'Keck_Pria_AS_Retest'
print 'label_name_list ', label_name_list
# specify dataset
k = 5
directory = '../../dataset/sample_fold/'
file_list = []
for i in range(k):
file_list.append('{}.csv'.format(i))
output_file_list = [directory + f_ for f_ in file_list]
print output_file_list[0:4]
train_pd = read_merged_data(output_file_list[0:4])
print output_file_list[4]
test_pd = read_merged_data([output_file_list[4]])
# extract data, and split training data into training and val
X_train, y_train = extract_feature_and_label(train_pd,
feature_name='Fingerprints',
label_name_list=label_name_list)
X_test, y_test = extract_feature_and_label(test_pd,
feature_name='Fingerprints',
label_name_list=label_name_list)
cross_validation_split = StratifiedShuffleSplit(y_train, 1, test_size=0.2, random_state=1)
for t_index, val_index in cross_validation_split:
X_t, X_val = X_train[t_index], X_train[val_index]
y_t, y_val = y_train[t_index], y_train[val_index]
print 'done data preparation'
task = RF_Tester()
task.get_rf(X_train, y_train, X_val, y_val, X_test, y_test)
return
def demo_single_classification():
with open(config_json_file, 'r') as f:
conf = json.load(f)
label_name_list = conf['label_name_list']
print 'label_name_list ', label_name_list
# specify dataset
k = 5
directory = '../../dataset/fixed_dataset/fold_{}/'.format(k)
file_list = []
for i in range(k):
file_list.append('file_{}.csv'.format(i))
# merge training and test dataset
dtype_list = {
'Molecule': np.str,
'SMILES': np.str,
'Fingerprints': np.str,
'Keck_Pria_AS_Retest': np.int64,
'Keck_Pria_FP_data': np.int64,
'Keck_Pria_Continuous': np.float64,
'Keck_RMI_cdd': np.float64
}
output_file_list = [directory + f_ for f_ in file_list]
print output_file_list[0:4]
train_pd = read_merged_data(output_file_list[0:4])
print output_file_list[4]
test_pd = read_merged_data([output_file_list[4]])
# extract data, and split training data into training and val
X_train, y_train = extract_feature_and_label(
train_pd, feature_name='Fingerprints', label_name_list=label_name_list)
X_test, y_test = extract_feature_and_label(test_pd,
feature_name='Fingerprints',
label_name_list=label_name_list)
cross_validation_split = StratifiedShuffleSplit(y_train,
1,
test_size=0.2,
random_state=1)
for t_index, val_index in cross_validation_split:
X_t, X_val = X_train[t_index], X_train[val_index]
y_t, y_val = y_train[t_index], y_train[val_index]
print 'done data preparation'
print conf['label_name_list']
task = SingleClassification(conf=conf)
task.train_and_predict(X_train, y_train, X_val, y_val, X_test, y_test,
PMTNN_weight_file)
store_data(transform_json_to_csv(config_json_file), config_csv_file)
return
dtype_list = {'Molecule': np.str,
'SMILES': np.str,
'Fingerprints': np.str,
'Keck_Pria_AS_Retest': np.int64,
'Keck_Pria_FP_data': np.int64,
'Keck_Pria_Continuous': np.float64,
'Keck_RMI_cdd': np.float64}
output_file_list = [directory + f_ for f_ in file_list]
print output_file_list[0:4]
train_pd = read_merged_data(output_file_list[0:4])
print output_file_list[4]
test_pd = read_merged_data([output_file_list[4]])
# extract data, and split training data into training and val
X_train, y_train = extract_feature_and_label(train_pd,
feature_name='Fingerprints',
label_name_list=label_name_list)
X_test, y_test = extract_feature_and_label(test_pd,
feature_name='Fingerprints',
label_name_list=label_name_list)
y_train_classification = reshape_data_into_2_dim(y_train[:, 0])
y_train_regression = reshape_data_into_2_dim(y_train[:, 1])
y_test_classification = reshape_data_into_2_dim(y_test[:, 0])
y_test_regression = reshape_data_into_2_dim(y_test[:, 1])
cross_validation_split = StratifiedShuffleSplit(y_train_classification, 1, test_size=0.2, random_state=1)
for t_index, val_index in cross_validation_split:
X_t, X_val = X_train[t_index], X_train[val_index]
y_t_classification, y_val_classification = y_train_classification[t_index], y_train_classification[val_index]
y_t_regression, y_val_regression = y_train_regression[t_index], y_train_regression[val_index]
def demo_multi_classification():
with open(config_json_file, 'r') as f:
conf = json.load(f)
label_name_list = conf['label_name_list']
print 'label_name_list ', label_name_list
# specify dataset
k = 5
directory = '../../dataset/keck_pcba/fold_{}/'.format(k)
file_list = []
for i in range(k):
file_list.append('file_{}.csv'.format(i))
output_file_list = [directory + f_ for f_ in file_list]
train_pd = read_merged_data(output_file_list[0:3])
train_pd.fillna(0, inplace=True)
val_pd = read_merged_data(output_file_list[3:4])
val_pd.fillna(0, inplace=True)
test_pd = read_merged_data(output_file_list[4:5])
test_pd.fillna(0, inplace=True)
multi_name_list = train_pd.columns[-128:].tolist()
multi_name_list.extend(label_name_list)
print 'multi_name_list ', multi_name_list
X_train, y_train = extract_feature_and_label(
train_pd, feature_name='Fingerprints', label_name_list=multi_name_list)
X_val, y_val = extract_feature_and_label(val_pd,
feature_name='Fingerprints',
label_name_list=multi_name_list)
X_test, y_test = extract_feature_and_label(test_pd,
feature_name='Fingerprints',
label_name_list=multi_name_list)
sample_weight_dir = '../../dataset/sample_weights/keck_pcba/fold_5/'
file_list = []
for i in range(k):
file_list.append('sample_weight_{}.csv'.format(i))
sample_weight_file = [sample_weight_dir + f_ for f_ in file_list]
sample_weight_pd = read_merged_data(sample_weight_file[0:3])
_, sample_weight = extract_feature_and_label(sample_weight_pd,
feature_name='Fingerprints',
label_name_list=labels_list)
print 'done data preparation'
task = MultiClassification(conf=conf)
task.train_and_predict(X_train,
y_train,
X_val,
y_val,
X_test,
y_test,
sample_weight=sample_weight,
PMTNN_weight_file=PMTNN_weight_file,
score_file=score_file)
store_data(transform_json_to_csv(config_json_file), config_csv_file)
whole_EF = []
for EF_ratio in task.EF_ratio_list:
EF_list = task.get_EF_score_with_existing_model(
X_test, y_test, PMTNN_weight_file, EF_ratio)
whole_EF.append([EF_ratio])
whole_EF.append(EF_list)
print(EF_ratio, EF_list)
print
print whole_EF
return