R/Pulldown_Visualization_Shiny.R

In dereksonderegger/PepSeq: Tools for analyzing Peptide Sequencing results

#' Create a Shiny application showing a scatterplot representing a pull-down
#'
#' The pulldown script created by John Altin creates a .csv file
#' with several columns that gives the number of reads for various
#' peptide sequences. This function reads the .csv file and creates
#' a Shiny application that allows the user to explore the pulldown.
#'
#' The input .csv file should have a columns denoting the protein name and location,
#' (by default we assume they are labeled `protein_ID` and `position`) and then
#' a bunch of response columns.  If the response column has a "XXX_cleaved" and
#' "XXX_uncleaved" pair, then we standardize the cleaved/uncleaved pair into a single
#' response.  If the column has no pair, then we don't do any standardization with it.
#'
#' @param input The input data frame  read in using the `import_pulldown()` function.
#'
#' @param height The height of the graph window (in pixels).
#'
#' @param peaks A logical flag denoting if the peaks should be highlighted
#'
#' @param peak_method Which method should be used for peak detection
#'
#' @param peak_param A parameter which controls the peak finding method. For PoT, it is the threshold.
#'
#' @param scales Contols if the y-scale should be 'fixed' across the rows or if the y-scales should be 'free'
#'              to vary across the different rows.
#'
#' @examples
#' file <- system.file("extdata", "example_counts.csv", package = "PepSeq")
#' plot_pulldown_Shiny(file, read_indicator='Y_')
#'
#' @export
plot_pulldown_Shiny <- function(input, height=400, peaks=FALSE, peak_method='PoT', peak_param=NA){

  df = input

  if( is.character(input) ){
    stop('input should be a data frame. Use import_pulldown() to load the data first!') }
  if( peaks & any(is.na(peak_param)) ){
    stop('peaks is TRUE, but peak_param is NA and there is no default value') }

  # figure out peaks
  if( peaks == TRUE ){
    Peak_Params <- df %>% group_by(Group) %>% count() %>% ungroup() %>% mutate(peak_param=peak_param) %>% select(-n)
    Peaks <- df %>%
      #mutate( signal = signal %>% pmax(signal,0) %>% as.integer() ) %>%
      left_join(Peak_Params, by='Group') %>%
      group_by(Group, protein_ID) %>%
      do( {identify_peaks_aux( .$position, .$signal, method='PoT', .$peak_param[1] ) }) %>%
      left_join( df, by=c('protein_ID', 'Group', 'Start'='position')) %>%
      rename( Start.index = index ) %>% select( Group, protein_ID, Peak, Start, End, Start.index) %>%
      left_join( df, by=c('protein_ID', 'Group', 'End'='position')) %>%
      rename( End.index = index ) %>% select( Group, protein_ID, Peak, Start, End, Start.index, End.index)
    # Print out the number of peaks
    Peaks %>% group_by(Group) %>% count() %>% rename(`Number of Peaks` = n) %>% print()
  }else{
    Peaks <- data.frame(Group=NULL, protein_ID=NULL, Peak=NULL, Start=NULL, End=NULL, Start.index=NULL, End.index=NULL)
  }


  Proteins <- df %>%
    group_by(protein_ID) %>%
    summarize(position = min(position),
              index = min(index))


  ymin=floor(   min(df$signal, na.rm=TRUE))
  ymax=ceiling( max(df$signal, na.rm=TRUE))
  n <- max(df$index)


  # create the UI
  ui <- fillPage(
    titlePanel("Peptide Sequence Exploration"),
    sidebarLayout(
      # Sidebar panel for inputs ----
      sidebarPanel(width=2,
                   numericInput(inputId = 'window_width', label = 'Window Width:',
                                value = 2000, min = 1, max=n, step = 100),
                   sliderInput(inputId = 'window_start', label = 'Window Start', min=1,
                               max=n, value = 1, step = 2000),
                   actionButton(inputId = 'window_previous', label = 'Previous'),
                   actionButton(inputId = 'window_next', label = 'Next'),
                   selectInput(inputId = 'protein', label  = 'Protein Selected',
                               choices = levels(df$protein_ID),
                               selected = Proteins %>% pull(protein_ID) %>%.[1] ),
                   numericInput(inputId = 'ymin', label = 'Minimum response:',
                                value = ymin, min = ymin, max=ymax),
                   numericInput(inputId = 'ymax', label = 'Maximum response:',
                                value = ymax, min = ymin, max=ymax),
                   selectInput(inputId = 'groups', label='Rows to View',
                               choices = unique(df$Group), selected=unique(df$Group),
                               multiple = TRUE)
      ),
      mainPanel(width=10,
                plotOutput(outputId = "Plot", height = "auto")
      )
    )
  )


  # define the server function
  server <- function(input, output, session) {
    # Update the controls as other controls get changed by the user
    # observeEvent( input$window_start, {
    #   # isolate(updateSelectInput(session, 'protein',
    #   #                   selected = df[input$window_start, 'protein_ID']))
    # })
    observeEvent( input$window_next, {
      updateSliderInput(session, 'window_start', value = input$window_start + input$window_width)
    })
    observeEvent( input$window_previous, {
      updateSliderInput(session, 'window_start', value = input$window_start - input$window_width)
    })
    observeEvent( input$protein, {
      isolate(updateSliderInput(session, 'window_start',
                        value=Proteins %>% filter(protein_ID == input$protein) %>% pull(index) ))
    })
    # observeEvent( input$window_width, {
    #   updateSliderInput( session, 'window_start', step = input$window_width)
    # })



    output$Plot <- renderPlot(
      {
        xmin <- round(input$window_start )
        xmax <- round(input$window_start + input$window_width)

        df.small <- df %>%
          filter( index > xmin, index < xmax ) %>%
          filter( Group %in% input$groups )

        P <- ggplot(df.small) +
          geom_point(size=.2, aes(x=position, y=signal)) +
          facet_grid( Group ~ protein_ID, scales='free_x', space='free_x') +
          coord_cartesian(ylim = c(input$ymin, input$ymax))

        if( peaks == TRUE & nrow(Peaks) >= 1 ){
          Peaks.small <- Peaks %>%
            filter( End.index > xmin, Start.index < xmax ) %>%
            filter( Group %in% input$groups )
          if(nrow(Peaks.small) > 0){
            P <- P +  geom_rect( data=Peaks.small, alpha=0.4, aes(xmin=Start, xmax=End, ymin=-Inf, ymax=Inf), fill='salmon')
          }
        }
        P
      },
      # height = function() { session$clientData$output_Plot_width }
      # height = input$plot_height
      # height = 400
      height = height
    )

  }

  shinyApp(ui = ui, server = server)

}