-
Notifications
You must be signed in to change notification settings - Fork 0
/
EMTAB_ENSEMBLE.R
362 lines (340 loc) · 15.4 KB
/
EMTAB_ENSEMBLE.R
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
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
library(gbm)
library(MLmetrics)
library(ModelMetrics)
library(caret)
library(plyr)
library(FCNN4R)
library(impute)
setwd("/Users/guanhaibin/Documents/BIORG/MM")
load("emc92_uams70.Rd")
genes_list<-unique(c(emc92,uams70))
eu_genes_list<-unique(genes_list)
E_genes_list<-read.csv("E_genes.csv",row.names = 1)
E_genes_list<-as.character(E_genes_list[,1])
model.entrezs<-c(eu_genes_list,E_genes_list)
model.entrezs<-unique(model.entrezs)
#EMTAB
Gene_expression=read.csv("EMTAB_balanced_Ensemble.csv",header=T,sep=",",row.names=1)
Gene_expression<-as.data.frame(t(Gene_expression))
HR_FLAG<-read.csv("EMTAB_target.csv",header=T,sep=",")
HR_FLAG<-t(HR_FLAG)
train_data<-merge(Gene_expression,HR_FLAG,by="row.names",all.x=TRUE)
row.names(train_data)<-train_data[,1]
train_data<-train_data[-c(1)]
colnames(train_data)[ncol(train_data)]<-c("label")
train_data$label<-as.factor(train_data$label)
rownames(train_data)<-gsub("X","E",rownames(train_data))
EMTAB_label<-train_data$label
model.entrezs<-intersect(colnames(Gene_expression),model.entrezs)
select.data<-setNames(as.data.frame(matrix(0,nrow=nrow(train_data),ncol=length(model.entrezs))),model.entrezs)
idx<-colnames(train_data)[colnames(train_data)%in%colnames(select.data)]
select.data[,idx]<-train_data[,idx]
rownames(select.data)<-rownames(train_data)
train_data <- select.data
train_data$label<-EMTAB_label
E_train_MM<-read.csv("E_train_MM.csv",row.names = 1)
E_train_MM<-as.character(E_train_MM[,1])
E_test_MM<-read.csv("E_test_MM.csv",row.names = 1)
E_test_MM<-as.character(E_test_MM[,1])
train_MM<-subset(train_data,rownames(train_data)%in%E_train_MM)
test_MM<-subset(train_data,rownames(train_data)%in%E_test_MM)
train_MM_label<-train_MM$label
train_MM$label<-NULL
test_MM_label<-test_MM$label
test_MM$label<-NULL
E_mean<-rowMeans(as.matrix(t(train_MM)))
E_sd<-rowSds(as.matrix(t(train_MM)))
train=sweep(train_MM,MARGIN=2,E_mean,"-")
train_MM<-sweep(train,MARGIN=2,E_sd,"/")
test=sweep(test_MM,MARGIN=2,E_mean,"-")
test_MM<-sweep(test,MARGIN=2,E_sd,"/")
train_MM$label<-train_MM_label
test_MM$label<-test_MM_label
nfea<-ncol(train_MM)-1
saveRDS(E_mean,"./E_mean.rds")
saveRDS(E_sd,"./E_sd.rds")
##create a dataframe to store prediction score for different model
ae_predict<-setNames(as.data.frame(matrix(ncol=9,nrow = 0)),c("model","Accuracy","AUC","F1_score","PRAUC","MCC","sensitivity","specific","classloss"))
class_result<-as.data.frame(matrix(ncol=nrow(test_MM)+1,nrow = 0))
pred_score<-as.data.frame(matrix(ncol=nrow(test_MM)+1,nrow = 0))
#function change label to 0 and 1 for calculate AUC and MMC
adjustAUC<-function(predict,truth){
{for (i in 1:length(predict)) {if (predict[i]==-1 ){{predict[i]=0}}}}
{for (i in 1:length(truth)) {if (truth[i]==-1 ){{truth[i]=0}}}}
return(AUC(predict,truth))
}
adjustmcc<-function(truth,predict,cutoff=1){
{for (i in 1:length(predict)) {if (predict[i]==-1 ){{predict[i]=0}}}}
{for (i in 1:length(truth)) {if (truth[i]==-1 ){{truth[i]=0}}}}
return(mcc(truth,predict,cutoff=1))
}
##SVM
##the random number seed is set before each algorithm is trained to ensure that each algorithm gets the same data partitions and repeats
library(e1071)
tc<-tune.control(cross=10)
svmtrain<-train_MM
svmfit<-svm(as.factor(label)~.,data=svmtrain,gamma=0.001,tune.control=tc,kernel = 'radial',probability=TRUE)
svm_pred_train<-predict(svmfit,svmtrain[,1:nfea])
prob<-predict(svmfit,test_MM[,1:nfea],probability=TRUE)
pred<-predict(svmfit,test_MM[,1:nfea])
pred<-as.vector(pred)
test_MM$label<-as.vector(test_MM$label)
SVM_accuracy<-Accuracy(pred,test_MM$label)
SVM_AUC<-adjustAUC(as.numeric(pred),as.numeric(test_MM$label))
SVM_F1<-F1_Score(as.numeric(test_MM$label),as.numeric(pred))
SVM_PRAUC<-PRAUC(as.numeric(pred),as.numeric(test_MM$label))
SVM_MCC<-adjustmcc(as.numeric(test_MM$label),as.numeric(pred),cutoff=1)
SVM_sensitivity<-Sensitivity(test_MM$label,pred)
SVM_specific<-Specificity(test_MM$label,pred)
SVM_classloss<-ZeroOneLoss(test_MM$label,pred)
#add model svm prediction score to class_result and ae_predict
asvm<-as.data.frame(t(c(SVM_accuracy,SVM_AUC,SVM_F1,SVM_PRAUC,SVM_MCC,SVM_sensitivity,SVM_specific,SVM_classloss)))
asvm<-as.data.frame(round(asvm,4))
asvm.model<-paste("SVM",sep="")
asvm<-cbind(asvm.model,asvm)
colnames(asvm)<-colnames(ae_predict)
ae_predict<-rbind(ae_predict,asvm)
pred.model<-paste("SVM",sep="")
pred_class<-cbind(pred.model,t(pred))
colnames(pred_class)<-colnames(class_result)
class_result<-rbind(class_result,pred_class)
##random forest
set.seed(7)
library(randomForest)
fitcontrol<-trainControl(method="cv",number=10,classProbs = TRUE)
RFtrain<-train_MM
levels(RFtrain$label)<-make.names(levels(factor(RFtrain$label)))
rffit<-caret::train(label~.,
RFtrain,
method="parRF",
tuneGrid=expand.grid(
.mtry=c(30,20,15,10,5,2,1)),
metric="Accuracy",
trControl=fitcontrol)
prob_rf<-predict(rffit,test_MM[,1:nfea],type='prob')
th<-0.5
pred_rf<-factor(ifelse(prob_rf$X1>th,"1","-1"))
rf_prob_train<-predict(rffit,RFtrain[,1:nfea],type="prob")
rf_pred_train<-factor(ifelse(rf_prob_train$X1>th,"1","-1"))
prob_rf<-predict(rffit,test_MM,type='prob')
pred_rf<-as.vector(pred_rf)
test_MM$label<-as.vector(test_MM$label)
rf_accuracy<-Accuracy(pred_rf,test_MM$label)
rf_AUC<-adjustAUC(as.numeric(pred_rf),as.numeric(test_MM$label))
rf_F1<-F1_Score(as.numeric(test_MM$label),as.numeric(pred_rf))
rf_PRAUC<-PRAUC(as.numeric(pred_rf),as.numeric(test_MM$label))
rf_MCC<-adjustmcc(as.numeric(test_MM$label),as.numeric(pred_rf),cutoff=1)
rf_sensitivity<-Sensitivity(test_MM$label,pred_rf)
rf_specific<-Specificity(test_MM$label,pred_rf)
rf_classloss<-ZeroOneLoss(test_MM$label,pred_rf)
#add model rf prediction score to class_result and ae_predict
arf<-as.data.frame(t(c(rf_accuracy,rf_AUC,rf_F1,rf_PRAUC,rf_MCC,rf_sensitivity,rf_specific,rf_classloss)))
arf<-as.data.frame(round(arf,4))
arf.model<-paste("RF",sep="")
arf<-cbind(arf.model,arf)
colnames(arf)<-colnames(ae_predict)
ae_predict<-rbind(ae_predict,arf)
pred_rf.model<-paste("RF",sep="")
pred_rf_class<-cbind(pred_rf.model,t(pred_rf))
colnames(pred_rf_class)<-colnames(class_result)
class_result<-rbind(class_result,pred_rf_class)
#Neural Network
nncontrol=trainControl(method = "cv", number = 10)
NNtrain<-train_MM
nn_Grid<-expand.grid(
.size=c(50,5),
.decay=0.00147)
nn_fit<-caret::train(label~.,
NNtrain,
method="nnet",
metric="Accuracy",
tuneGrid=nn_Grid,
preProcess = c('center', 'scale'),
MaxNWts=10000,
maxit=100,
trace=FALSE)
pred_nn<-predict(nn_fit,test_MM[,1:nfea])
nn_pred_train<-predict(nn_fit,RFtrain[,1:nfea])
prob_nn<-predict(nn_fit,test_MM,type='prob')
nn_prob_train<-predict(nn_fit,RFtrain[,1:nfea],type="prob")
nn_pred_train<-predict(nn_fit,RFtrain[,1:nfea])
#prob_nn<-predict(nn_fit,test_MM,type='prob')
pred_nn<-as.vector(pred_nn)
test_MM$label<-as.vector(test_MM$label)
nn_accuracy<-Accuracy(pred_nn,test_MM$label)
nn_AUC<-adjustAUC(as.numeric(pred_nn),as.numeric(test_MM$label))
nn_F1<-F1_Score(as.numeric(test_MM$label),as.numeric(pred_nn))
nn_PRAUC<-PRAUC(as.numeric(pred_nn),as.numeric(test_MM$label))
nn_MCC<-adjustmcc(as.numeric(test_MM$label),as.numeric(pred_nn),cutoff=1)
nn_sensitivity<-Sensitivity(test_MM$label,pred_nn)
nn_specific<-Specificity(test_MM$label,pred_nn)
nn_classloss<-ZeroOneLoss(test_MM$label,pred_nn)
#add model rf prediction score to class_result and ae_predict
a_nn<-as.data.frame(t(c(nn_accuracy,nn_AUC,nn_F1,nn_PRAUC,nn_MCC,nn_sensitivity,nn_specific,nn_classloss)))
a_nn<-as.data.frame(round(a_nn,4))
a_nn.model<-paste("ANN",sep="")
a_nn<-cbind(a_nn.model,a_nn)
colnames(a_nn)<-colnames(ae_predict)
ae_predict<-rbind(ae_predict,a_nn)
pred_nn.model<-paste("ANN",sep="")
pred_nn_class<-cbind(pred_nn.model,t(pred_nn))
colnames(pred_nn_class)<-colnames(class_result)
class_result<-rbind(class_result,pred_nn_class)
#Learning Vector Quantization
set.seed(7)
lvqcontrol<-trainControl(method = "cv", number = 10)
lvq_grid <- expand.grid(size=c(5,10,20,50), k=c(2,4,5,7,10))
LVQtrain<-train_MM
lvq_fit<-caret::train(label~.,
LVQtrain,
method="lvq",
trControl=lvqcontrol,
tuneGrid=lvq_grid)
pred_lvq<-predict(lvq_fit,test_MM[,1:nfea])
lvq_pred_train<-predict(lvq_fit,RFtrain[,1:nfea])
pred_lvq<-as.vector(pred_lvq)
test_MM$label<-as.vector(test_MM$label)
lvq_accuracy<-Accuracy(pred_lvq,test_MM$label)
lvq_AUC<-adjustAUC(pred_lvq,as.numeric(test_MM$label))
lvq_F1<-F1_Score(as.numeric(test_MM$label),as.numeric(pred_lvq))
lvq_PRAUC<-PRAUC(as.numeric(pred_lvq),as.numeric(test_MM$label))
lvq_MCC<-adjustmcc(as.numeric(test_MM$label),as.numeric(pred_lvq),cutoff=1)
lvq_sensitivity<-Sensitivity(test_MM$label,pred_lvq)
lvq_specific<-Specificity(test_MM$label,pred_lvq)
lvq_classloss<-ZeroOneLoss(test_MM$label,pred_lvq)
#add model lvq prediction score to class_result and ae_predict
a_lvq<-as.data.frame(t(c(lvq_accuracy,lvq_AUC,lvq_F1,lvq_PRAUC,lvq_MCC,lvq_sensitivity,lvq_specific,lvq_classloss)))
a_lvq<-as.data.frame(round(a_lvq,4))
a_lvq.model<-paste("lvq",sep="")
a_lvq<-cbind(a_lvq.model,a_lvq)
colnames(a_lvq)<-colnames(ae_predict)
ae_predict<-rbind(ae_predict,a_lvq)
pred_lvq.model<-paste("lvq",sep="")
pred_lvq_class<-cbind(pred_lvq.model,t(pred_lvq))
colnames(pred_lvq_class)<-colnames(class_result)
class_result<-rbind(class_result,pred_lvq_class)
##GBM
set.seed(7)
gbm_fit<-caret::train(label~.,
RFtrain,
method="gbm",
verbose=FALSE,
trControl=fitcontrol)
prob_gbm<-predict(gbm_fit,test_MM[,1:nfea],type='prob')
th<-0.5
pred_gbm<-factor(ifelse(prob_gbm$X1>th,"1","-1"))
gbm_prob_train<-predict(gbm_fit,RFtrain[,1:nfea],type="prob")
gbm_pred_train<-factor(ifelse(gbm_prob_train$X1>th,"1","-1"))
pred_gbm<-as.vector(pred_gbm)
test_MM$label<-as.vector(test_MM$label)
gbm_accuracy<-Accuracy(pred_gbm,test_MM$label)
gbm_AUC<-adjustAUC(pred_gbm,as.numeric(test_MM$label))
gbm_F1<-F1_Score(as.numeric(test_MM$label),as.numeric(pred_gbm))
gbm_PRAUC<-PRAUC(as.numeric(pred_gbm),as.numeric(test_MM$label))
gbm_MCC<-adjustmcc(as.numeric(test_MM$label),as.numeric(pred_gbm),cutoff=1)
gbm_sensitivity<-Sensitivity(test_MM$label,pred_gbm)
gbm_specific<-Specificity(test_MM$label,pred_gbm)
gbm_classloss<-ZeroOneLoss(test_MM$label,pred_gbm)
#add model gbm prediction score to class_result and ae_predict
a_gbm<-as.data.frame(t(c(gbm_accuracy,gbm_AUC,gbm_F1,gbm_PRAUC,gbm_MCC,gbm_sensitivity,gbm_specific,gbm_classloss)))
a_gbm<-as.data.frame(round(a_gbm,4))
a_gbm.model<-paste("GBM",sep="")
a_gbm<-cbind(a_gbm.model,a_gbm)
colnames(a_gbm)<-colnames(ae_predict)
ae_predict<-rbind(ae_predict,a_gbm)
pred_gbm.model<-paste("GBM",sep="")
pred_gbm_class<-cbind(pred_gbm.model,t(pred_gbm))
colnames(pred_gbm_class)<-colnames(class_result)
class_result<-rbind(class_result,pred_gbm_class)
##glmnet Generalized Linear Model
set.seed(7)
gridsearch_for_lambda = data.frame (alpha = 0,
lambda = c (2^c(-15:15), 3^c(-15:15)))
train_control = trainControl (method="cv", number =10,
savePredictions =TRUE , allowParallel = FALSE,classProbs = TRUE )
glmfit<-caret::train(label~.,
RFtrain,
method="glmnet",
tuneGrid=gridsearch_for_lambda,
trControl=train_control,
preProcess=NULL)
prob_glm<-predict(glmfit,test_MM[,1:nfea],type='prob')
th<-0.5
pred_glm<-factor(ifelse(prob_glm$X1>th,"1","-1"))
glm_prob_train<-predict(glmfit,RFtrain[,1:nfea],type="prob")
glm_pred_train<-factor(ifelse(glm_prob_train$X1>th,"1","-1"))
pred_glm<-as.vector(pred_glm)
test_MM$label<-as.vector(test_MM$label)
glm_accuracy<-Accuracy(pred_glm,test_MM$label)
glm_AUC<-adjustAUC(pred_glm,as.numeric(test_MM$label))
glm_F1<-F1_Score(as.numeric(test_MM$label),as.numeric(pred_glm))
glm_PRAUC<-PRAUC(as.numeric(pred_glm),as.numeric(test_MM$label))
glm_MCC<-adjustmcc(as.numeric(test_MM$label),as.numeric(pred_glm),cutoff=1)
glm_sensitivity<-Sensitivity(test_MM$label,pred_glm)
glm_specific<-Specificity(test_MM$label,pred_glm)
glm_classloss<-ZeroOneLoss(test_MM$label,pred_glm)
#add model rf prediction score to class_result and ae_predict
a_glm<-as.data.frame(t(c(glm_accuracy,glm_AUC,glm_F1,glm_PRAUC,glm_MCC,glm_sensitivity,glm_specific,glm_classloss)))
a_glm<-as.data.frame(round(a_glm,4))
a_glm.model<-paste("GLM",sep="")
a_glm<-cbind(a_glm.model,a_glm)
colnames(a_glm)<-colnames(ae_predict)
ae_predict<-rbind(ae_predict,a_glm)
pred_glm.model<-paste("GLM",sep="")
pred_glm_class<-cbind(pred_glm.model,t(pred_glm))
colnames(pred_glm_class)<-colnames(class_result)
class_result<-rbind(class_result,pred_glm_class)
##Stacked model
#combine all the predictions of train data from above classifiers to train stacked model
stacked_tc<-trainControl(method = 'cv',number=10,classProbs = TRUE)
train_label<-RFtrain$label
combo_train<-data.frame(svm_pred_train,rf_pred_train,nn_pred_train,lvq_pred_train,gbm_pred_train,glm_pred_train,train_label)
levels(combo_train$train_label)<-make.names(levels(factor(combo_train$train_label)))
label<-as.factor(test_MM$label)
combo_test<-data.frame(pred,pred_rf,pred_nn,pred_lvq,pred_gbm,pred_glm,label)
colnames(combo_test)<-colnames(combo_train)
levels(combo_test$train_label)<-make.names(levels(factor(combo_test$train_label)))
fit_stacked<-caret::train(as.factor(train_label)~.,
combo_train,
method="nnet",
metric="Accuracy",
tuneGrid=nn_Grid,
MaxNWts=10000,
maxit=100,
trace=FALSE)
prob_stacked<-predict(fit_stacked,combo_test[,1:7],type='prob')
th<-0.5
pred_stacked<-factor(ifelse(prob_stacked$X1>th,"1","-1"))
pred_stacked<-as.vector(pred_stacked)
test_MM$label<-as.vector(test_MM$label)
stacked_accuracy<-Accuracy(pred_stacked,test_MM$label)
stacked_AUC<-adjustAUC(pred_stacked,as.numeric(test_MM$label))
stacked_F1<-F1_Score(as.numeric(test_MM$label),as.numeric(pred_stacked))
stacked_PRAUC<-PRAUC(as.numeric(pred_stacked),as.numeric(test_MM$label))
stacked_MCC<-adjustmcc(as.numeric(test_MM$label),as.numeric(pred_stacked),cutoff=1)
stacked_sensitivity<-Sensitivity(test_MM$label,pred_stacked)
stacked_specific<-Specificity(test_MM$label,pred_stacked)
stacked_classloss<-ZeroOneLoss(test_MM$label,pred_stacked)
#add model rf prediction score to class_result and ae_predict
a_stacked<-as.data.frame(t(c(stacked_accuracy,stacked_AUC,stacked_F1,stacked_PRAUC,stacked_MCC,stacked_sensitivity,stacked_specific,stacked_classloss)))
a_stacked<-as.data.frame(round(a_stacked,4))
a_stacked.model<-paste("stacked_model",sep="")
a_stacked<-cbind(a_stacked.model,a_stacked)
colnames(a_stacked)<-colnames(ae_predict)
ae_predict<-rbind(ae_predict,a_stacked)
pred_stacked.model<-paste("stacked_model",sep="")
pred_stacked_class<-cbind(pred_stacked.model,t(pred_stacked))
colnames(pred_stacked_class)<-colnames(class_result)
class_result<-rbind(class_result,pred_stacked_class)
write.csv(class_result,file="EMATB_TRAIN_pred_HR_FLAG.csv")
write.csv(ae_predict,file="EMTAB__TRAIN_ccuracy.csv")
#save model seperately
saveRDS(svmfit,"./E_svmfit.rds")
saveRDS(rffit,"./E_rffit.rds")
saveRDS(nn_fit,"./E_nn_fit.rds")
saveRDS(lvq_fit,"./E_lvq_fit.rds")
saveRDS(gbm_fit,"./E_gbm_fit.rds")
saveRDS(glmfit,"./E_glmfit.rds")
saveRDS(fit_stacked,"./E_fit_stacked.rds")
E_genes_eu<-model.entrezs
write.csv(E_genes_eu,"E_genes_eu.csv")