DBSCAN clustering in GUI or online software

Question

[rob.costa1234]

Is there any web based or GUI based tool that has Density-based spatial clustering (DBSCAN) r package option for clustering. It has to be recommended to a non bioinformatic colleague to explore. Thanks

Answer 1

[Jean-Karim Heriche]

Why not use a Shiny app? You can use the code below as starting point.

library(shiny)
library(dbscan)
library(ggplot2)
library(shinycssloaders) # Loading/busy animations
ui <- fluidPage(
  titlePanel("Input", windowTitle = "DBSCAN"),
  ## Sidebar with input and output definitions
  sidebarLayout(
    position = c("left", "right"),
    fluid = TRUE,
    ## Sidebar panel for inputs
    sidebarPanel(
      width = 3,
      h4("Data file"),
      ## Input: Select a csv file
      fileInput("datafile", "Select data file (in csv format)",
                multiple = FALSE,
                accept = c("text/csv",
                           "text/comma-separated-values,text/plain",
                           ".csv")),
      ## Checkbox if file has header
      checkboxInput("header", "File has header", TRUE),
      ## Select separator
      radioButtons("sep", "Separator",
                   choices = c(Comma = ",",
                               Semicolon = ";",
                               Tab = "\t"),
                   selected = "\t"),
      ## Select quotes
      radioButtons("quote", "Quote",
                   choices = c(None = "",
                               "Double Quote" = '"',
                               "Single Quote" = "'"),
                   selected = '"'),
      tags$hr(style="border-top: 1px solid #000000;"),        
      h4("Plot"),
      ## Select variables to display
      uiOutput("plotVariables"),
      tags$hr(style="border-top: 1px solid #000000;"),
      h4("DBSCAN"),
      ## Select variables to use for clustering
      uiOutput("clusterVariables"),
      ## DBSCAN input parameters
      numericInput('eps', 'Size of the epsilon neighbourhood (eps)', NULL),
      numericInput('minPts', 'Number of minimum points in the eps region (minPts)', 5, min = 1),
      actionButton("clusterButton", "Cluster away!")      
    ),
    mainPanel(
      width = 9,
      fluidRow( 
        column(9,
               fluidRow(
                 h3('Plot'),
                 withSpinner(
                   plotOutput('plot', height = 600, width = '100%')
                 )
               )
        )
      )
    )    
  )
)
server <- function(input, output, session) {
  rv <- reactiveValues(data = NULL, clusters = NULL) 
  ## Read uploaded data file
  observeEvent(input$datafile, {   
    if (is.null(input$datafile)) return(NULL)
    req(input$datafile) ## Proceed only if file selected     
    rv$data <- read.csv(input$datafile$datapath,
                            header = input$header,
                            sep = input$sep,
                            quote = input$quote,
                            stringsAsFactors = FALSE)    
      ## Remove non-numeric columns to be safe for clustering
      rv$data <- rv$data[,sapply(rv$data, is.numeric)]
  })
  ## Dynamically generate UI input to select variables once data is uploaded and read
  ## Do once to select variables to plot and once to select variables to use for clustering
  output$plotVariables <- renderUI({
    selectizeInput(inputId = "plotVar", 
                   label = "Variables to plot",
                   multiple = TRUE,
                   choices = names(rv$data),
                   options = list(maxItems = 2))
  })
  output$clusterVariables <- renderUI({
    selectizeInput(inputId = "clusterVar", 
                   label = "Variables to use for clustering",
                   multiple = TRUE,
                   choices = names(rv$data))
  })
  ## Do clustering once button is pressed
  observeEvent(input$clusterButton, {
    if(!is.na(input$eps && length(input$clusterVar)>0)) {
     dbr <- dbscan(as.matrix(rv$data[, input$clusterVar]), input$eps, input$minPts)
     rv$clusters <- dbr$cluster
    }
  }) 
  ## Plot
  output$plot <- renderPlot({
    selectedCols <- input$plotVar
    if (length(selectedCols) == 2) {
      p <- ggplot(data = rv$data, aes_string(x = selectedCols[1], y = selectedCols[2])) +
        labs(x = eval(selectedCols[1]), y = eval(selectedCols[2]))
      if(!is.null(rv$clusters)) { # We have clusters
        p <- p +  geom_point(aes(colour = factor(rv$clusters))) +
        scale_colour_brewer('Clusters', palette = 'Set1')
      } else { # No clustering
        p <- p + geom_point(color = 'darkblue')
      }
      p <- p + coord_cartesian(expand = TRUE)
      plot(p)
    }
  })
}
## Run the application
shinyApp(ui = ui, server = server)

[Iut Tri]

What is the sample data to run this GUI? should there be a rule? I managed to run tis GUI but when entering data can't run it. Thank you very much.