-
Notifications
You must be signed in to change notification settings - Fork 0
/
cnn_validate3_onehot.py
267 lines (202 loc) · 9 KB
/
cnn_validate3_onehot.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
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
#https://www.dataquest.io/blog/learning-curves-machine-learning/
import pandas as pd
import numpy as np
import random
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelBinarizer
from sklearn.preprocessing import MinMaxScaler
scaler = MinMaxScaler()
def multiclass_roc_auc_score(y_test, y_pred, average="macro"):
lb = LabelBinarizer()
lb.fit(y_test)
y_test = lb.transform(y_test)
y_pred = lb.transform(y_pred)
return roc_auc_score(y_test, y_pred, average=average)
df = pd.read_csv("train_dataset.csv")
df = df.dropna() # To drop Null values
print(df.shape)
df = df.drop_duplicates('sgRNA_sequence', keep='last')
print("&&&&&&&&")
print(df.shape)
print("&&&&&&&&")
print(df.head())
#########################################################
target_value_list=list(df.iloc[:, 12])
print("22222222222", len(target_value_list))
print(target_value_list[0])
target_value_class=[]
for i in target_value_list:
if type(i)!=str:
if i>=0 and i<0.052:
i=0#'Min_Q1'
target_value_class.append(i)
elif i>=0.052 and i<0.88:
i=1#'Q1_Q2'
target_value_class.append(i)
elif i>=0.88 and i<1.63:
i=2#'Q2_Q3'
target_value_class.append(i)
else:
i=3#'Q3_Max'
target_value_class.append(i)
print("length of target_value_class",len(target_value_class))
print(target_value_class[0:50])
df['label']=target_value_class
print(df.head())
######################################################################
print("#######################")
df2 = pd.read_csv("test_dataset.csv")
df2 = df2.dropna() # To drop Null values
print(df2.shape)
df2 = df2.drop_duplicates('sgRNA_sequence', keep='last')
print("&&&&&&&&")
print(df2.shape)
print("&&&&&&&&")
print(df2.head())
#########################################################
target_value_list=list(df2.iloc[:, 12])
print("22222222222", len(target_value_list))
print(target_value_list[0])
target_value_class=[]
for i in target_value_list:
if type(i)!=str:
if i>=0 and i<0.031:
i=0#'Min_Q1'
target_value_class.append(i)
elif i>=0.031 and i<0.84:
i=1#'Q1_Q2'
target_value_class.append(i)
elif i>=0.84 and i<1.59:
i=2#'Q2_Q3'
target_value_class.append(i)
else:
i=3#'Q3_Max'
target_value_class.append(i)
print("length of target_value_class",len(target_value_class))
print(target_value_class[0:50])
df2['label']=target_value_class
print(df2.head())
##################################################################
#combine onehot_encoding of train and test
union_reference_kmer_set=set(df.iloc[:, 1]).union(set(df2.iloc[:, 1]))
union=list(union_reference_kmer_set)
print(len(union))
df['sgRNA_sequence']=pd.Categorical(df['sgRNA_sequence'], categories=list(union))
df2['sgRNA_sequence']=pd.Categorical(df2['sgRNA_sequence'], categories=list(union))
####################################################################
####################################################################
features_train=['Efficiency','Specificity','BS_Length','PEEK','GWAS','Distance_exon_BS (D)','sgRNA_Conc_Q0','sgRNA_Conc_T5_K562_Rep1_normalized']
#X_train = df[features_train]
#print(X_train.shape)
#X_train= scaler.fit_transform(X_train)
X_train = df[features_train]
#insert onehot encoding of reference-kmer
Onehot=pd.get_dummies(df['sgRNA_sequence'], prefix='sgRNA_sequence')
X_train= pd.concat([X_train,Onehot],axis=1)
a,b=X_train.shape
print("b=",b)
features_test=['Efficiency','Specificity','BS_Length','PEEK','GWAS','Distance_exon_BS (D)','sgRNA_Conc_Q0','sgRNA_Conc_T5_K562_Rep2_normalized']
#X_test = df2[features_test]
#print(X_test.shape)
#X_test= scaler.fit_transform(X_test)
X_test = df2[features_test]
#insert onehot encoding of reference-kmer
Onehot=pd.get_dummies(df2['sgRNA_sequence'], prefix='sgRNA_sequence')
X_test= pd.concat([X_test,Onehot],axis=1)
c,d=X_test.shape
print("d=",d)
X_train= scaler.fit_transform(X_train)
X_test= scaler.fit_transform(X_test)
y_train = df['label']
print(y_train.head())
y_test = df2['label']
print(y_test.head())
#######################################################
X_train = np.expand_dims(np.random.normal(size=(a, b)),axis=-1) #shap input-train to have 3 D tensor in order to be processed with conv1D
# Ytrain => [213412, 10]
#y_train = np.random.choice([0,1,2,3], size=(a,1))y_train = to_categorical(y_train)
X_test = np.expand_dims(np.random.normal(size=(c, d)),axis=-1) #shap input-test to have 3 D tensor in order to be processed with conv1D
# Ytrain => [213412, 10]
#y_test = np.random.choice([0,1,2,3], size=(c,1))
####################################################################
#Sequence/signal classification with 1D convolutions: https://keras.io/getting-started/sequential-model-guide/ , https://blog.goodaudience.com/introduction-to-1d-convolutional-neural-networks-in-keras-for-time-sequences-3a7ff801a2cf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Dropout,Flatten,BatchNormalization,LeakyReLU
from tensorflow.keras.layers import Embedding
from tensorflow.keras.layers import Conv1D, GlobalAveragePooling1D, MaxPooling1D,GlobalMaxPooling1D
from tensorflow.keras.optimizers import SGD, Adam,RMSprop
from tensorflow.keras.utils import to_categorical
#one_hot_label = to_cateorical(input_labels)
from tensorflow.keras.callbacks import TensorBoard, EarlyStopping,ModelCheckpoint,ReduceLROnPlateau, Callback
import matplotlib.pyplot as plt
from sklearn import metrics #Import scikit-learn metrics module for accuracy calculation
from sklearn.metrics import classification_report #for classifier evaluation
from sklearn.metrics import roc_auc_score # for printing AUC
y_train = to_categorical(y_train)
y_test = to_categorical(y_test)
#Implement cnn model4 for training embedding layer, available at https://machinelearningmastery.com/develop-word-embedding-model-predicting-movie-review-sentiment/
learning_rate=0.0001
#################
model = Sequential()
model.add(Conv1D(filters=1024, kernel_size=3, activation='relu',input_shape=(b, 1))) #or kernel_size=8 and MaxPooling1D(pool_size=1)
model.add(MaxPooling1D(pool_size=2))
model.add(Conv1D(filters=1024, kernel_size=3, activation='relu'))
model.add(Flatten())
#model.add(Dense(10, activation='relu'))
model.add(Dense(4, activation='softmax'))
#########
print(model.summary())
callbacks_list = [ModelCheckpoint(filepath='best_model.{epoch:02d}-{val_loss:.2f}.h5',monitor='val_loss', save_best_only=True),EarlyStopping(monitor='acc', patience=1)]
model.compile(loss='categorical_crossentropy',
optimizer=Adam(lr=learning_rate), metrics=['accuracy'])
BATCH_SIZE = 256
EPOCHS = 5
history = model.fit(X_train,
y_train,
batch_size=BATCH_SIZE,
epochs=EPOCHS,
callbacks=callbacks_list,
validation_split=0.1,
verbose=1)
# evaluate the cnn model on the same dataset
_, accuracy = model.evaluate(X_test, y_test, verbose=0)
#_, accuracy = model.evaluate(X_test, y_test, verbose=1)
print('Accuracy: %.2f' % (accuracy*100))
loss_train, accuracy_train = model.evaluate(X_train, y_train, verbose=1)
loss_test, accuracy_test = model.evaluate(X_test, y_test, verbose=1)
print('Accuracy on training: %.2f' % (accuracy_train*100))
print('Accuracy on testing: %.2f' % (accuracy_test*100))
#print('Test loss:', loss_test)
train_loss = history.history['loss']
val_loss = history.history['val_loss']
train_acc = history.history['accuracy']
val_acc = history.history['val_accuracy']
xc = range(50)
#plot epochs versus accuracy curve.
plt.plot(train_acc, label="Training accuracy")
plt.plot(val_acc, label="Validation accuracy")
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.legend(loc="best")
plt.show()
plt.plot(train_loss, label="Training loss")
plt.plot(val_loss, label="Validation loss")
plt.xlabel('Epochs')
plt.ylabel('loss')
plt.legend(loc="best")
plt.show()
# make class predictions with the model
#y_pred = model.predict_classes(X_test)
y_pred =np.argmax(model.predict(X_test), axis=-1)
import numpy as np #to handle the error Classification metrics can't handle a mix of multilabel-indicator and multiclass targets
y_test=np.argmax(y_test, axis=1)
y_test[1]
print(y_test[1])
print(classification_report(y_test, y_pred))
print(y_test.shape)
print(y_pred.shape)
y_test=y_test.reshape(-1,1) #reshape from 1 column array to two column array, because some np functions does not like missing dimensions (-,)
y_pred=y_pred.reshape(-1,1)
print(y_test.shape)
print(y_pred.shape)
print("AUC=",multiclass_roc_auc_score(y_test,y_pred)) #https://medium.com/@plog397/auc-roc-curve-scoring-function-for-multi-class-classification-9822871a6659