如何在R Shiny中结合checkboxGroupInput与其他输入绘制散点图

First, let's fix up your incomplete sample data setup so we have a working foundation:

library(dplyr)
library(permute)
library(Rtsne)
library(shiny)
library(ggplot2)
library(stringr)

set.seed(1)
meta.df <- data.frame(gene_id=paste0("id",1:10),symbol=paste0("n",rep(permute::shuffle(5),2)),stringsAsFactors=F)
clusters.df <- data.frame(cell=paste0("c",1:100),cluster=rep(permute::shuffle(10),10),sample=paste0("s",rep(permute::shuffle(5),20)),stringsAsFactors=F)
mat <- matrix(rnorm(10*100),10,100,dimnames=list(meta.df$gene_id,clusters.df$cell))

# Generate complete t-SNE output and merge with metadata
tsne.obj <- Rtsne::Rtsne(t(mat), perplexity = 10, seed = 1)
tsne.df <- as.data.frame(tsne.obj$Y) %>%
  rename(TSNE1 = V1, TSNE2 = V2) %>%
  bind_cols(clusters.df) %>%
  # Add gene expression values per cell
  left_join(t(mat) %>% as.data.frame() %>% rownames_to_column("cell"), by = "cell")

Now let's build a Shiny app that seamlessly combines checkboxGroupInput (for selecting genes to visualize) with other interactive inputs to create dynamic t-SNE plots.

Full Working Shiny App Code #

ui <- fluidPage(
  titlePanel("Interactive t-SNE Scatter Plot"),
  
  sidebarLayout(
    sidebarPanel(
      # Checkbox group for gene selection
      checkboxGroupInput(
        inputId = "selected_genes",
        label = "Select Genes to Visualize",
        choices = meta.df$gene_id,
        selected = meta.df$gene_id[1]  # Default to first gene
      ),
      
      # Additional input: choose categorical variable to color points by
      selectInput(
        inputId = "color_by",
        label = "Color Points By",
        choices = c("Cluster" = "cluster", "Sample" = "sample"),
        selected = "cluster"
      ),
      
      # Optional: pick how to display gene expression
      radioButtons(
        inputId = "expr_display",
        label = "Show Gene Expression As",
        choices = c("Point Size" = "size", "Point Opacity" = "alpha"),
        selected = "size"
      )
    ),
    
    mainPanel(
      plotOutput("tsne_plot")
    )
  )
)

server <- function(input, output) {
  
  # Reactive data handler: updates based on user inputs
  reactive_plot_data <- reactive({
    req(input$selected_genes)  # Wait until genes are selected
    
    # Calculate mean expression if multiple genes are chosen
    if(length(input$selected_genes) > 1) {
      tsne.df %>%
        mutate(expression = rowMeans(select(., all_of(input$selected_genes)))) %>%
        select(TSNE1, TSNE2, all_of(input$color_by), expression)
    } else {
      tsne.df %>%
        select(TSNE1, TSNE2, all_of(input$color_by), expression = all_of(input$selected_genes))
    }
  })
  
  # Render the dynamic t-SNE plot
  output$tsne_plot <- renderPlot({
    plot_data <- reactive_plot_data()
    color_col <- input$color_by
    display_type <- input$expr_display
    
    # Base plot structure
    p <- ggplot(plot_data, aes(x = TSNE1, y = TSNE2)) +
      geom_point(aes(color = .data[[color_col]]), size = 2) +
      labs(
        title = "t-SNE Cell Visualization",
        x = "TSNE1",
        y = "TSNE2",
        color = str_to_title(color_col)
      ) +
      theme_minimal()
    
    # Add gene expression visualization based on user choice
    if(display_type == "size") {
      p <- p + geom_point(aes(size = expression), alpha = 0.7) +
        labs(size = ifelse(length(input$selected_genes) >1, "Mean Expression", "Gene Expression"))
    } else {
      p <- p + geom_point(aes(alpha = expression), size = 2) +
        labs(alpha = ifelse(length(input$selected_genes) >1, "Mean Expression", "Gene Expression"))
    }
    
    p
  })
}

shinyApp(ui = ui, server = server)

Key Details Explained #

• Checkbox Group Integration: Users can select one or more genes. If multiple are chosen, the app calculates the mean expression across those genes per cell for cleaner visualization.
• Input Combination: The app pairs gene selection with:
  • A dropdown to choose between coloring points by cluster or sample
  • Radio buttons to decide if gene expression is shown via point size or opacity
• Reactive Updates: All inputs trigger real-time updates to the plot, thanks to reactive data handling.
• Flexibility: You can easily extend this by adding more inputs (like a slider for point size, or a dropdown for specific gene symbols instead of IDs) based on your needs.

Just make sure you have all required packages installed before running the app!

内容的提问来源于stack exchange，提问作者dan