#################################### Normalizing the RNA-seq data and analyzing differentially expressed genes (DEGs) ####################################
foldChange=2
padj=0.01
library("edgeR")
rt1 <- read.table("sampleExp.txt", header=T, sep="\t", row.names=1, check.names=F)
head(rt1)

rt2 <- read.table("sample.txt", header=T, sep="\t")
head(rt2)

newdf1 <- as.data.frame(t(rt1))
newdf1[, "id"] <- rownames(newdf1)
newdf2 <- rt2
newdf3 <- merge(newdf2, newdf1, by="id", all.x=TRUE)
rt <- newdf3
rt[, 1:4]

rownames(rt)=rt[,1]
exp=t(rt[,4:ncol(rt)])
dimnames=list(rownames(exp),colnames(exp))
data=matrix(as.numeric(as.matrix(exp)),nrow=nrow(exp),dimnames=dimnames)
data=avereps(data)
data=data[rowMeans(data)>1,]
nrow(data)
ncol(data)

group=rt[, 2]                         
block=rt[, 3]
design <- model.matrix(~block+group)
y <- DGEList(counts=data)
y <- calcNormFactors(y)
y <- estimateCommonDisp(y)
y <- estimateTagwiseDisp(y)
y <- estimateDisp(y,design)
fit <- glmQLFit(y, design)
qlf <- glmQLFTest(fit)
et <- topTags(qlf, n=nrow(data))
nrow(et)
ordered_tags <- et

allDiff=ordered_tags$table
allDiff=allDiff[is.na(allDiff$FDR)==FALSE,]
diff=allDiff
newData=y$pseudo.counts


write.table(diff,file="edgerOut.xls",sep="\t",quote=F)
diffSig = diff[(diff$FDR < padj & (diff$logFC>foldChange | diff$logFC<(-foldChange))),]
write.table(diffSig, file="diffSig.xls",sep="\t",quote=F)
diffUp = diff[(diff$FDR < padj & (diff$logFC>foldChange)),]
write.table(diffUp, file="up.xls",sep="\t",quote=F)
diffDown = diff[(diff$FDR < padj & (diff$logFC<(-foldChange))),]
write.table(diffDown, file="down.xls",sep="\t",quote=F)

normalizeExp=rbind(id=colnames(newData),newData)
write.table(normalizeExp,file="normalizeExp.txt",sep="\t",quote=F,col.names=F)  
newData1 <- as.data.frame(t(newData))
newData1[, "id"] <- rownames(newData1)
newData2 <- merge(rt[newData1[, "id"], 1:3], newData1, by="id", all.y=TRUE)
write.table(newData2, "TCGA.txt", row.names=FALSE, sep="\t", quote=FALSE)

diffExp=rbind(id=colnames(newData),newData[rownames(diffSig),])
write.table(diffExp,file="diffmRNAExp.txt",sep="\t",quote=F,col.names=F) 

heatmapData <- newData[rownames(diffSig),]




#################################### Plotting heatmap ####################################
data <- read.table("diffmRNAExp.txt", header=T, sep="\t", row.names=1, check.names=F)
data <- as.matrix(log10(data+0.001))
data <- data[1:nrow(data),]

library(pheatmap)
annotation <- read.table("sample.txt",header=T, sep="\t", row.names=1)
p <- pheatmap(
    data, annotation=annotation,
    color=colorRampPalette(c("navy", "white", "firebrick3"))(50),
    scale="row", clustering_distance_row="correlation",
    fontsize=9, fontsize_row=6, show_rownames=F, show_colnames=F)

pdf("heatmap.pdf")
p
dev.off()

#################################### Removing batch effects and normalizing the microarray data ####################################
library(sva)
library(limma)
rt=read.table("merge.txt",sep="\t",header=T,check.names=F)
rt=as.matrix(rt)
rownames(rt)=rt[,1]
exp=rt[,2:ncol(rt)]
dimnames=list(rownames(exp),colnames(exp))
data=matrix(as.numeric(as.matrix(exp)),nrow=nrow(exp),dimnames=dimnames)
batchType=c(rep(1,155),rep(2,167))
modType=c(rep("normal",77),rep("tumor",78),rep("normal",52),rep("tumor",115))
mod = model.matrix(~as.factor(modType))
outTab=ComBat(data, batchType, mod, par.prior=TRUE)
outTab=rbind(geneNames=colnames(outTab),outTab)
write.table(outTab,file="normalize.txt",sep="\t",quote=F,col.names=F)



#################################### Visualizing the ceRNA network ####################################
df <- read.table("network.txt",sep = "\t",row.names = 1,header = T)
head(df)
library(ggalluvial)
mycol <- rep(c("#223D6C","#D20A13","#FFD121","#088247","#11AA4D","#58CDD9","#7A142C","#5D90BA","#029149","#431A3D","#91612D","#6E568C","#E0367A","#D8D155","#64495D","#7CC767",
"#223D6C","#D20A13","#FFD121","#088247","#11AA4D","#58CDD9","#7A142C","#5D90BA","#029149","#431A3D","#91612D","#6E568C","#E0367A","#D8D155","#64495D","#7CC767",
"#223D6C","#D20A13","#FFD121","#088247","#11AA4D","#58CDD9","#7A142C","#5D90BA","#029149","#431A3D","#91612D","#223D6C","#D20A13","#FFD121","#088247","#11AA4D","#58CDD9","#7A142C","#5D90BA","#029149","#431A3D","#91612D","#6E568C","#E0367A","#D8D155","#64495D","#7CC767"
,"#223D6C","#D20A13","#FFD121","#088247","#11AA4D","#58CDD9","#7A142C","#5D90BA","#029149","#431A3D","#91612D","#6E568C","#E0367A","#D8D155","#64495D","#7CC767"),3)

UCB_lodes <- to_lodes_form(df[,1:ncol(df)],
                           axes = 1:ncol(df),
                           id = "Cohort")
dim(UCB_lodes)

head(UCB_lodes)
tail(UCB_lodes)

ggplot(UCB_lodes,
       aes(x = x, stratum = stratum, alluvium = Cohort,
           fill = stratum, label = stratum)) +
  scale_x_discrete(expand = c(0, 0)) + 
  geom_flow(width = 1/5) +
  geom_stratum(alpha = .9,width = 1/7) +
  geom_text(stat = "stratum", size = 1.5,color="black") +
  

  scale_fill_manual(values = mycol) +

  xlab("") + ylab("") +
  theme_bw() +
  theme(panel.grid =element_blank()) +
  theme(panel.border = element_blank()) +
  theme(axis.line = element_blank(),axis.ticks = element_blank(),axis.text = element_blank()) +
  ggtitle("")+
  guides(fill = FALSE) 

ggsave("sankey3.pdf")



#################################### Survival analysis ####################################
options(stringsAsFactors=FALSE)
library(survival)
library(survminer)
outTab=data.frame()

picDir="picture"
dir.create(picDir)

library(survival)
library(qvalue)
rt=read.table("tumor.time.txt",header=T,sep="\t",row.names=1,check.names=F)
rt[,"futime"]=rt[,"futime"]/365

rt[1:4,1:4]
rt1=log2(rt[,3:ncol(rt)]+1)
rt=cbind(rt[,1:2],rt1)
rt[1:4,1:4]
sur.cut <- surv_cutpoint(rt, time="futime", event="fustat",
                         variables=colnames(rt[,3:ncol(rt)]))
summary(sur.cut)
labels <- paste(c(">", "<="),
                rep(summary(sur.cut)[,"cutpoint"], each=2), sep="")
for(i in colnames(rt[,3:ncol(rt)])){
    outfile <- paste("picture/cox_cutoff_", i, ".jpg", sep="")
    jpeg(outfile, width=20, height=20, units="cm", res=350)
    print(plot(sur.cut, i, palette="npg"))
    dev.off()

   outPdf=paste("picture/cox_cutoff", i,".pdf", sep="")
   pdf(file=outPdf, onefile=FALSE)
   print(plot(sur.cut, i, palette="npg"))
   dev.off()
}
oldnames <- colnames(rt)
colnames(rt) <- gsub("-", ".", colnames(rt))
sur.cut <- surv_cutpoint(rt, time="futime", event="fustat",
                         variables=colnames(rt[,3:ncol(rt)]))
summary(sur.cut)
rt <- surv_categorize(sur.cut, variables=colnames(rt[,3:ncol(rt)]))
head(rt)
colnames(rt) <- oldnames


outlst <- list()
for(i in colnames(rt[,3:ncol(rt)])){
 cox <- coxph(Surv(futime, fustat) ~ rt[, i], data=rt)
 coxSummary = summary(cox)
 aoe <- cbind(as.data.frame(summary(cox)[["coefficients"]]),as.data.frame(summary(cox)[["conf.int"]]))
 abc <- t(aoe)[,1]
 coxP=coxSummary$coefficients[,"Pr(>|z|)"]

   rt1=rt[rt[, i] == "high",]
   rt2=rt[rt[, i] == "low",]
   n1=nrow(rt1)
   n2=nrow(rt2)
   surTab1=summary(survfit(Surv(futime, fustat) ~ 1, data = rt1))
   surTab2=summary(survfit(Surv(futime, fustat) ~ 1, data = rt2))
   medianTab1=surTab1$table
   medianTab2=surTab2$table
   
   model <- survdiff(Surv(futime, fustat) ~ rt[, i], data = rt)
   chisq <- model[["chisq"]]
   df <- length(model[["n"]]) - 1
   pvalue <- pchisq(chisq, df, lower.tail=FALSE)

   label <- labels[rep(colnames(rt[,3:ncol(rt)]), each=2) == i]
   outlst[[i]] <- cbind(
       data.frame(var=i, level=c("high", "low"), label=label),
       rbind(medianTab1, medianTab2), pvalue_logrank=pvalue,
       hr=1/abc[6], lcl=1/abc[9], ucl=1/abc[8])

   diff=survdiff(Surv(futime, fustat) ~ rt[,i],data = rt)
   fit <- survfit(Surv(futime, fustat) ~ rt[,i], data = rt)
   pValue=1-pchisq(diff$chisq, df=1)
   outTab=rbind(outTab,cbind(gene=i,coxSummary$coefficients,coxSummary$conf.int,KM=pValue,
   H_med=medianTab1["median"],H_0.95LCL=medianTab1["0.95LCL"],H_0.95UCL=medianTab1["0.95UCL"],
   L_med=medianTab2["median"],L_0.95LCL=medianTab2["0.95LCL"],L_0.95UCL=medianTab2["0.95UCL"]))
   pval=0
   if(pValue<0.05){
     pval=signif(pValue,4)
     pval=format(pval, scientific = TRUE)
     }else{
     pval=round(pValue,3)
     }
   if(pValue<0.05){
   HR <- paste("Hazard Ratio = ", round(1/abc[6],2), sep = "")
   CI <- paste("95% CI: ", paste(round(1/abc[9],2), round(1/abc[8],2), sep = " - "), sep = "")
   fig <- ggsurvplot(fit, data=rt, 
          ncensor.plot = TRUE,
          censor = T,
          pval = paste(pval = ifelse(pvalue < 0.001, "p < 0.001", 
                       paste("p = ",round(pvalue,3), sep = "")),
                        HR, CI, sep = "\n"),
          conf.int=TRUE, 
          conf.int.style = "ribbon",
          legend.title=i, legend.labs=label, 
          palette = c("#AC88FF", "#FC717F"), 
          xlab = "Years after surgery (year) ",ylab = "Overall survival rate",
          ggtheme=theme(
            legend.key=element_rect(fill="white", colour=NA),
            panel.border=element_rect(fill="transparent", colour="black"),
            panel.background=element_rect(fill="white")
            ))
   geneName=unlist(strsplit(i,"\\|",))[1]
   tiffFile=paste(geneName,".survival.tiff",sep="")
   outTiff=paste(picDir,tiffFile,sep="\\")
   tiff(file=outTiff,width = 15,height = 15,units ="cm",compression="lzw",bg="white",res=600)
   print(fig)
   dev.off()

   outPdf=paste("picture/", geneName,".survival.pdf",sep="")
   pdf(file=outPdf, onefile=FALSE)
   print(fig)
   dev.off()
   }
}
outdf <- do.call(rbind, outlst)
write.csv(outdf, "surv_median_logrank.csv", row.names=FALSE, na="")