Search code examples
ggplot2plotshinypcascatter

How to create a R shiny app for getting PCA plot

I am just starting to learn R shiny and am trying to create a shiny app that produces scatter plot for principal component analysis and allows user to choose various principal components on the X and Y axis of scatter plot. I know how to write R code to do PCA but I just cant seem to get the shiny app to get me what I need.. I have tried following the examples available for Iris kmeans clustering but I am having trouble getting the scatter plot. Here is my code so far (P.S. my original dataset has genes as rows and samples as columns (columns 1 through 10 are cancer samples, 11 through 20 are normal):

data<-read.table("genes_data.txt", header=TRUE, row.names=1)
pca_data<-prcomp(t(data), scale=T)
summary(pca_data)
pca_sig.var<-pca_data$sdev^2
pca_sig.var.per<-round(pca_sig.var/sum(pca_sig.var)*100, 1)

pca_sig.data<-data.frame(Sample=rownames(pca_data$x), PC1=pca_data$x[,1], PC2=pca_data$x[,2], PC3=pca_data$x[,3], PC4=pca_data$x[,4], PC5=pca_data$x[,5])
pca_sig.data<-pca_sig.data[-1]

pca_sig.data2<-pca_sig.data
pca_sig.data2$category=rep("CANCER", 20)
pca_sig.data2$category[11:20]=rep("NORMAL", 10)
View(pca_sig.data2)
ggplot(data=pca_sig.data2, aes(x=PC1, y=PC2, label=category, colour=category))+
  geom_point(size=2, stroke=1, alpha=0.8, aes(color=category))+
  xlab(paste("PCA1 - ", pca_sig.var.per[1], "%", sep=""))+
  ylab(paste("PCA2 - ", pca_sig.var.per[2], "%", sep=""))+
  theme_bw()+
  ggtitle("My PCA Graph")


ui<-pageWithSidebar(
  headerPanel('Gene Data PCA'),
  sidebarPanel(
    selectInput('xcol', 'X Variable', names(pca_sig.data2[,1:5])),
    selectInput('ycol', 'Y Variable', names(pca_sig.data2[,1:5]),
                selected=names(pca_sig.data2)[[2]])

  ),
  mainPanel(
    plotOutput('plot1')
  )
)


server<- function(input, output, session) {

  # Combine the selected variables into a new data frame
  selectedData <- reactive({
    pca_sig.data2[, c(input$xcol, input$ycol)]
  })

  output$plot1 <- renderPlot({
    palette(c("#E41A1C", "#377EB8"))

    par(mar = c(5.1, 4.1, 0, 1))
    plot(selectedData(),
         col=selectedData()$category,
         pch = 20, cex = 3)
    points(selectedData()[,1:5], pch = 4, cex = 4, lwd = 4)
  })

}
shinyApp(ui = ui, server = server)

At the end, when I run the app, I get "Error:undefined columns selected"

Also, for simplicity sake let's assume that my original dataset that I want to do PCA on looks something like this (in reality I have about 600 genes and 20 samples):

probeID<-c("gene1", "gene2", "gene3", "gene4","gene5")

BCR1<-c(28.005966, 30.806433, 17.341375, 17.40666, 30.039436)
BCR2<-c(30.973469, 29.236025, 30.41161, 20.914383, 20.904331)
BCR3<-c(26.322796, 25.542833, 22.460772, 19.972183, 30.409641)
BCR4<-c(26.441898, 25.837685, 23.158352, 20.379173, 33.81327)
BCR5<-c(39.750206, 19.901133, 28.180124, 22.668673, 25.748884)
CTL6<-c(23.004385, 28.472675, 23.81621, 26.433413, 28.851719)
CTL7<-c(22.239546, 28.741674, 23.754929, 26.015385, 28.16368)
CTL8<-c(29.590443, 30.041988, 21.323061, 24.272501, 18.099016)
CTL9<-c(15.856442, 22.64224, 29.629637, 25.374926, 22.356894)
CTL10<-c(38.137985, 24.753338, 26.986668, 24.578161, 19.223558)
data<-data.frame(probeID, BCR1, BCR2, BCR3, BCR4, BCR5, CTL6, CTL7, CTL8, CTL9, CTL10)

where BCR1 through BCR5 are the cancer samples and CTL6 through CTL10 are the normal samples.

Solution

  • Is this what you want?

    server<- function(input, output, session) {

    # Combine the selected variables into a new data frame
    selectedData <- reactive({
        pca_sig.data2[c(input$xcol, input$ycol, 'category')]
    })

    output$plot1 <- renderPlot({
        palette(c("#E41A1C", "#377EB8"))

        plot(selectedData()[,c(1:2)], col=factor(selectedData()$category), pch = 20, cex = 3)
        points(selectedData()[,c(1:2)], pch = 4, cex = 4, lwd = 4)
    })

}

    The result is like this: enter image description here