-
Notifications
You must be signed in to change notification settings - Fork 2
/
dense4.py
129 lines (103 loc) · 5.06 KB
/
dense4.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
import keras as ks
from keras import Model
import pandas as pd
from sklearn.model_selection import train_test_split
from keras.layers import Input, Dense,Dropout,BatchNormalization
from sklearn.model_selection import KFold
from keras import regularizers
from keras.optimizers import SGD
from keras.constraints import UnitNorm
import os
import numpy as np
def get_threshold_metrics(y_true, y_pred, drop_intermediate=False,
disease='all'):
"""
Retrieve true/false positive rates and auroc/aupr for class predictions
Arguments:
y_true - an array of gold standard mutation status
y_pred - an array of predicted mutation status
disease - a string that includes the corresponding TCGA study acronym
Output:
dict of AUROC, AUPR, pandas dataframes of ROC and PR data, and cancer-type
"""
import pandas as pd
from sklearn.metrics import roc_auc_score, roc_curve
from sklearn.metrics import precision_recall_curve, average_precision_score
roc_columns = ['fpr', 'tpr', 'threshold']
pr_columns = ['precision', 'recall', 'threshold']
if drop_intermediate:
roc_items = zip(roc_columns,
roc_curve(y_true, y_pred, drop_intermediate=False))
else:
roc_items = zip(roc_columns, roc_curve(y_true, y_pred))
roc_df = pd.DataFrame.from_dict(dict(roc_items))
prec, rec, thresh = precision_recall_curve(y_true, y_pred)
pr_df = pd.DataFrame.from_records([prec, rec]).T
pr_df = pd.concat([pr_df, pd.Series(thresh)], ignore_index=True, axis=1)
pr_df.columns = pr_columns
auroc = roc_auc_score(y_true, y_pred, average='weighted')
aupr = average_precision_score(y_true, y_pred, average='weighted')
return {'auroc': auroc, 'aupr': aupr, 'roc_df': roc_df,
'pr_df': pr_df, 'disease': disease}
gpu_id = '0,1'
os.environ['CUDA_VISIBLE_DEVICES'] = str(gpu_id)
os.system('echo $CUDA_VISIBLE_DEVICES')
ks.backend.clear_session()
print('reading')
x_np=np.array(pd.read_csv('genefile.csv',index_col=0,header=0))
y_df=np.array(pd.read_csv('statefile.csv',index_col=0,header=0))
strat=pd.read_csv('stratfile.csv',index_col=0,header=0)
print('read down')
input_dim=2048
x_train_all,x_test_all, y_train_all, y_test_all =train_test_split(x_df.iloc[:,:input_dim],y_df,test_size=0.1, random_state=0,stratify=strat)#,stratify=strat)
input = Input(shape=(input_dim,))
dense_1 = Dense(200, init='normal',activation='relu')(input)
bn_2=BatchNormalization()(dense_1)
dropout_1=Dropout(0.5)(bn_2)
dense_2 = Dense(100,init='normal', activation='relu')(dropout_1)
bn_3=BatchNormalization()(dense_2)
dropout_2=Dropout(0.5)(bn_3)
dense_4 = Dense(30,init='normal', activation='relu')(dropout_2)
bn_4=BatchNormalization()(dense_4)
dropout_3=Dropout(0.5)(bn_4)
dense_3 = Dense(10,init='normal', activation='relu')(dropout_3) # ,
final_1 = Dense(1, activation='sigmoid')(dense_3)
model = Model(inputs=input, outputs=final_1)
model.compile(optimizer='sgd',
loss='binary_crossentropy',#loss='binary_crossentropy',
metrics=['accuracy'])
print(model.summary())
seed = 1
np.random.seed(seed)
all_specificity = list()
all_precision = list()
all_sensitivity = list()
accurancy_sum = 0
loss_sum = 0
k_fold = KFold(5,True, random_state=1)
index = k_fold.split(X=x_train_all, y=y_train_all)
index_train_all=np.array(x_train_all.index)
cv_results_df=list()
for train_index, test_index in index:
X_train = np.array(x_train_all.iloc[train_index,:])
X_cv = np.array(x_train_all.iloc[test_index,:])
y_train = np.array(y_train_all.iloc[train_index,:])
y_cv = np.array(y_train_all.iloc[test_index,:])
test_index=np.array(index_train_all[test_index])
print("Training --------------------")
learning_rate_reduction = ks.callbacks.ReduceLROnPlateau(monitor='loss', patience=5, verbose=1,
epsilon=0.0003, factor=0.5, min_lr=1e-5)#epsilon=0.0003, factor=0.9, min_lr=0.00001)
t = model.fit(X_train, y_train, batch_size=128, epochs=50, verbose=1,
callbacks=[learning_rate_reduction], validation_data=(X_cv,y_cv)) # batch_size=128ds
print("\nTesting --------------------")
cv_pred=np.concatenate((model.predict(X_cv),y_cv),axis=1)
cv_df=pd.DataFrame(data=cv_pred, columns=['dignosis','total_status'],index=test_index)
cv_results_df.append(cv_df)
cv_df=pd.concat(cv_results_df,axis=0)
cv_df.to_csv('DENSE4cv_results.csv')
train_pred=np.concatenate((model.predict(np.array(x_train_all.loc[:,:])),np.array(y_train_all.loc[:,:])),axis=1)
train_df=pd.DataFrame(data=train_pred, columns=['dignosis','total_status'],index=x_train_all.index)
train_df.to_csv('DENSE4train_results.csv')
test_pred=np.concatenate((model.predict(np.array(x_test_all.loc[:,:])),np.array(y_test_all.loc[:,:])),axis=1)
test_df=pd.DataFrame(data=test_pred, columns=['dignosis','total_status'],index=x_test_all.index)
test_df.to_csv('DENSE4test_results.csv')