-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathfig3_LASSO_SVM.R
139 lines (139 loc) · 6.11 KB
/
fig3_LASSO_SVM.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
load("IDCSCmatrix.Rda")
load("IDCSCmeta.Rda")
rownames(matrix3)<-matrix3$circRNA
matrix3$circRNA<-NULL
quantile(rowSums(as.matrix(matrix3[1:584])> 0),0.6)
quantile(rowSums(as.matrix(matrix3[1:584])> 0),0.7)
quantile(rowSums(as.matrix(matrix3[1:584])> 0),0.8)
quantile(rowSums(as.matrix(matrix3[1:584])> 0),0.9)
matrix3q7<-matrix3[which(rowSums(as.matrix(matrix3)> 0)>=4&rowSums(as.matrix(matrix3)> 0)<7),]
matrix3q8<-matrix3[which(rowSums(as.matrix(matrix3)> 0)>=7&rowSums(as.matrix(matrix3)> 0)<20),]
matrix3q9<-matrix3[which(rowSums(as.matrix(matrix3)> 0)>=20),]
################################################################
###1 ML data preparation
matrix3t<-as.data.frame(t(matrix3q9)) #according to which quantile to be investigated
Samples<-as.data.frame(rownames(matrix3t))
colnames(Samples)<-"Samples"
meta5<-merge(Samples,meta4,all=F)
tag<-gsub("cancer","4",meta5$Type)
tag<-gsub("normal","2",tag)
matrix3t<-cbind(tag,matrix3t)
# partitioning into training (70%) and validation (30%)
set.seed(42)
# randomly sample 70% of the row IDs for training; the remaining 30% serve as validation
train.rows <- sample(rownames(matrix3t), dim(matrix3t)[1]*0.7)
# collect all the columns with training row ID into training set:
matrix3t.train <- matrix3t[train.rows, ]
# assign row IDs that are not already in the training set, into validation
valid.rows <- setdiff(rownames(matrix3t), train.rows)
matrix3t.valid <- matrix3t[valid.rows, ]
##################################################################
###2 feature selection by lasso
set.seed(42)
library(glmnet)
library(pROC)
lasso.results <- c()
for (j in 1:50) {
s <- matrix3t.train[sample(1:nrow(matrix3t.train),ceiling(dim(matrix3t.train)[1]*0.5),replace=F),]
ts <- apply(s,2,as.numeric)
y <- as.matrix(ts[,1])
x <- as.matrix(ts[,-1])
cv.fit <- cv.glmnet(x,y,family="binomial", type.measure = "auc", nfolds=5)
co<-coef(cv.fit,s="lambda.1se")
name <- rownames(co)[co[,1]!=0]
lasso.results <- c(lasso.results, name)
pred = predict(cv.fit, newx = x, type = 'response',s ="lambda.min")
roc<-roc(y,pred)
pdf(paste(j,".ROC01.pdf",sep=""))
plot.roc(roc,col="red",print.auc =TRUE,print.auc.col = "darkgreen",auc.polygon = TRUE,auc.polygon.col = "pink")
dev.off() }
plot(cv.fit)
f1 = glmnet(x, y, family="binomial",type.measure = "auc", nfolds=5)
plot(f1, xvar="lambda", label=TRUE)
freq.lasso.results <- as.data.frame(table(lasso.results))
freq.lasso.results<-freq.lasso.results[-1,]
write.csv(freq.lasso.results,"matrix3q7lasso.csv")
###################################################################
###3 SVM diganostic model construction
set.seed(42)
library(pROC)
library(caret)
library(e1071)
fitControl <- trainControl(method = "repeatedcv",
number = 5,
repeats = 10,
## Estimate class probabilities
classProbs = TRUE,
## Evaluate performance using
## the following function
summaryFunction = twoClassSummary)
matrix3t.train3$Y<-gsub("4","cancer",matrix3t.train3$Y)
matrix3t.train3$Y<-gsub("2","normal",matrix3t.train3$Y)
matrix3t.valid3$Y<-gsub("4","cancer",matrix3t.valid3$Y)
matrix3t.valid3$Y<-gsub("2","normal",matrix3t.valid3$Y)
svmFit <- train(Y ~ ., data=matrix3t.train3,
method = "svmRadial",
trControl = fitControl,
# preProc = c("center", "scale"),
tuneLength = 10,
metric = "ROC")
train.pred <- predict(svmFit,matrix3t.train3)
validation.pred <- predict(svmFit,matrix3t.valid3)
tpred <- predict(svmFit,matrix3t.train3, type = 'prob')
vpred <- predict(svmFit,matrix3t.valid3, type = 'prob')
train.con <- confusionMatrix(train.pred, factor(matrix3t.train3$Y))
valid.con <- confusionMatrix(validation.pred, factor(matrix3t.valid3$Y))
#testing
testing.pred <- predict(svmFit,HCCNP3)
testing.con <- confusionMatrix(testing.pred, factor(HCCNP3$Y))
spred <- predict(svmFit,HCCNP3, type = 'prob')
train.con
valid.con
testing.con
##################################################################
###4 plot ROC for SVM
z<-gsub("cancer","4",matrix3t.train3$Y)
z<-gsub("normal","2",z)
q<-gsub("cancer","4",matrix3t.valid3$Y)
q<-gsub("normal","2",q)
w<-gsub("cancer","4",HCCNP3$Y)
w<-gsub("normal","2",w)
roc1<-roc(q,vpred[,1])
plot.roc(roc1,col="red",print.auc =TRUE,print.auc.col = "red")
roc2<-roc(z,tpred[,1])
plot.roc(roc2,col="dodgerblue",print.auc =TRUE,print.auc.col = "dodgerblue",add=TRUE)
roc3<-roc(w,spred[,1])
plot.roc(roc3,col="darkgreen",print.auc =TRUE,print.auc.col = "darkgreen",add=TRUE)
#################################################################
###loading output of LASSO
lasso9<-read.csv("matrix3q9lasso.csv",header=T)
lasso8<-read.csv("matrix3q8lasso.csv",header=T)
lasso7<-read.csv("matrix3q7lasso.csv",header=T)
lasso9$X<-NULL
lasso8$X<-NULL
lasso7$X<-NULL
lasso9<-lasso9[which(lasso9$Freq>10),]
lasso8<-lasso8[which(lasso8$Freq>15),]
lasso7<-lasso7[which(lasso7$Freq>15),]
library("ggplot2")
ggplot(lasso7,aes(x= reorder(lasso.results,Freq), y=Freq,fill=Freq)) +
geom_bar(stat = "identity") +
coord_flip()+
scale_fill_gradient(low = "pink", high = "red")+
xlab("circRNAs") +
scale_y_continuous(name="Frequency",expand=c(0,0))+
theme_classic()
lasso9$Block<-"Top10"
lasso8$Block<-"Top1020"
lasso7$Block<-"Top2030"
lasso<-rbind(lasso9,lasso8,lasso7)
matrix3b<-matrix3[as.vector(lasso$lasso.results),]
library(pheatmap)
library(RColorBrewer)
anno2<-lasso[,c(1,3)]
rownames(anno2)<-anno2[,1]
anno2$lasso.results<-NULL
ann_colors = list(Block=c(Top10=brewer.pal(3,"Set3")[1],Top1020=brewer.pal(3,"Set3")[2],Top2030=brewer.pal(3,"Set3")[3]),Type=c(cancer="tomato",normal="dodgerblue"))
matrix3b<-as.data.frame(t(matrix3b))
pheatmap(matrix3b,scale='column',annotation_row=anno,annotation_col=anno2,annotation_colors =ann_colors,cluster_col=T,cluster_row=T,color = colorRampPalette(c("navy","white", "red" ))(100),cellwidth=3,cellheight=1,fontsize =1)
write.csv(colnames(matrix3b),"lassotopcircRNA.csv")