app.R

library(shiny)
library(plotly)
library(DT)
library(ggplot2)
library(tibble)
library(ggsurvfit)
library(survminer)
library(gridExtra)
library(tidyverse)
library(broom)
library(survival)
library(shinydashboard)

select=read.csv("data/subtype_select.csv")
  k=1
  m=select$cancer[k]
  n=select$subtype[k]
  
  
  
  data1=read.csv(paste( "data/",m,"_", n, ".csv",sep=""))
  data1<-data1[order(data1$BatchId),]
  
  data1<-data1[,-1]
  data=data1[data1$os_time!=0&!is.na(data1$os_time),]
  data=data[data$os_time>0,]
  names=colnames(data)[1883:ncol(data)]
  data2=data[,2:1882]
  data3<-log2(data2+1)
  genes<-colnames(data3)
  datause<-cbind(data3,data$os_status,data$os_time)
  colnames(datause)[1882]<-"os_status"
  colnames(datause)[1883]<-"os_time"
  
  datausepre=datause[,1:1881]
  geneid=colnames(datausepre)
  exclude=c()
  means1=colMeans(datausepre)
  ps_rho_cutoff=0.9
  thr=2
  x_sub=datausepre[,means1>thr]
  xcorr=cor(x_sub)
  
  for(i in 1:ncol(x_sub)){
    for(j in i:ncol(x_sub)){
      if(xcorr[i,j]>ps_rho_cutoff & xcorr[i,j]!=1){
        exclude_1=ifelse(means1[geneid==rownames(xcorr)[i]]>=means1[geneid==colnames(xcorr)[j]],
                         colnames(xcorr)[j], rownames(xcorr)[i])
        exclude=c(exclude, exclude_1)
      }
    }
  }
  if(length(unique(exclude))==0){
    datapre=x_sub
  }else{
    datapre=x_sub[,!(colnames(x_sub) %in% unique(exclude))]
  }
  
  geneidfinal=colnames(datapre)
  
  #data.train=cbind(data.train.raw$submitter_id,x.train,data.train.raw[,1883:ncol(data.train.raw)])
  datausefinal=cbind(datause$os_time,datause$os_status,datapre)
  colnames(datausefinal)[1]='os_time'
  
  colnames(datausefinal)[2]='os_status'
  
  
  
  #dataplot=read.csv(paste('/users/yilin/Downloads/cureitsurvival-2/cure',m,n,'.csv'))
  dataplot=read.csv(paste('data/cure',m,n,'.csv'))
  dataplot$color[dataplot$scoretest<0.01]=1
  dataplot$color[dataplot$scoretest>=0.01]=0
  dataplot$color=as.factor(dataplot$color)
  dataplot$curepvalue<-apply(dataplot[,c('curepvalue1','curepvalue2','curepvalue3')],1,min)
  dataplot<-dataplot%>%filter(curepvalue!=0)
  dataplot$curepvalue=-log10(dataplot$curepvalue)
  dataplot$survivalpvalue<-apply(dataplot[,c('survivalpvalue1','survivalpvalue2','survivalpvalue3')],1,min)
  dataplot<-dataplot%>%filter(survivalpvalue!=0)
  dataplot$survivalpvalue=-log10(dataplot$survivalpvalue)
  dataplot$coxpvalue=-log10(dataplot$scoretest)
  
  
  dataplot<-dataplot[!dataplot$survivalpvalue>=10,]
  dataplot<-dataplot[!dataplot$curepvalue>=10,]
  dataplot<-dataplot[!dataplot$coxpvalue>=10,]
  #dataplot<-dataplot[!abs(dataplot$survival)>=10,]
  #dataplot<-dataplot[!abs(dataplot$cure)>=10,]
  #dataplot<-dataplot[!abs(dataplot$cox)>=10,]
  
  
  
  ui <- dashboardPage(
    dashboardHeader(title = "Basic dashboard"),
    dashboardSidebar(),
    dashboardBody(
      fluidPage(
        fluidRow(
          column(6,plotlyOutput("plot")),
          column(6,plotlyOutput("text"),plotOutput("text1"))
        )
        #plotlyOutput("plot"),
        #plotlyOutput("text"),
        #plotlyOutput("text1")
      )
    )
  )
  server<-function(input,output){
    
    #data <- reactive({
    #dataplot
    # })
    
    output$plot <- renderPlotly({
      #d <- data()
      plot_ly(dataplot, x= ~curepvalue, y=~survivalpvalue, split=~color, mode = "markers", type = "scatter", source="mysource")
      #ggplot(data=dataplot,aes(x=curepvalue,y=survivalpvalue,col=color))+geom_point()+scale_color_manual(values=c('Black','Red'))
    })
    
    
    
    
    output$text <- renderPlotly({
      event.data <- event_data("plotly_click", source = "mysource")
      print(event.data)
      if(is.null(event.data)) { return(NULL)}
      index=which(dataplot$survivalpvalue==event.data$y)
      geneid=dataplot$gene[index]
      datausefinal$label1=cut(datausefinal[,geneid],breaks=c(min(datausefinal[,geneid]),quantile(datausefinal[,geneid],0.25),quantile(datausefinal[,geneid],0.5),quantile(datausefinal[,geneid],0.75),max(datausefinal[,geneid])),
                              include.lowest = T,right = F,label=c(0,1,2,3))
      fit<-survfit(Surv(datausefinal$os_time,datausefinal$os_status)~datausefinal$label1,data=datausefinal)
      #plot_ly(datausefinal,x=~os_time,y=~os_status)
      #ggplot(data=datausefinal,aes(x=os_time,y=os_status))+geom_point()
      p<-ggsurvplot(fit,data=datausefinal, risk.table = TRUE, risk.table.y.text.col = TRUE)+labs(title=geneid)
      ggplotly(p[[1]])
    })
    
    
    output$text1 <- renderPlot({
      event.data <- event_data("plotly_click", source = "mysource")
      print(event.data)
      if(is.null(event.data)) { return(NULL)}
      index=which(dataplot$survivalpvalue==event.data$y)
      geneid=dataplot$gene[index]
      #plot_ly(data=datausefinal,x=~str(geneid),type='histogram')
      hist(datausefinal[,geneid],main=geneid)
    })
    
    output$download_plotly_widget <- downloadHandler(
      filename = function() {
        paste("/users/yilin/Downloads/data-", Sys.Date(), ".html", sep = "")
      },
      content = function(file) {
        # export plotly html widget as a temp file to download.
        saveWidget(as_widget(session_store$plt), file, selfcontained = TRUE)
      }
    )
    
  }
  
  shinyApp(ui, server)