R/cytomapperShiny.R

In cytomapper: Visualization of highly multiplexed imaging data in R

#'Shiny application to visualise gated cells on images
#'
#'This shiny application allows users to gate cells based on their raw or
#'transformed expression values and visualises gated cells on their
#'corresponding images.
#'
#'@param object a \code{\linkS4class{SingleCellExperiment}} object.
#'@param mask (optional) a \code{\linkS4class{CytoImageList}} containing
#'    single-channel \code{\linkS4class{Image}} objects.
#'@param image (optional) a \code{\linkS4class{CytoImageList}} object containing
#'    single or multi-channel \code{\linkS4class{Image}} objects.
#'@param cell_id character specifying the \code{colData(object)} entry, in which
#'    the integer cell IDs are stored. These IDs should match the integer pixel
#'    values in the segmentation mask object (\code{mask}).
#'@param img_id character specifying the \code{colData(object)} and
#'    \code{mcols(mask)} and/or \code{mcols(image)} entry, 
#'    in which the image IDs are stored.
#'@param ... parameters passed to the \code{\link{plotCells}} or
#'    \code{\link{plotPixels}} function.
#'
#'@section User inputs:
#'This function requires at least a \code{\linkS4class{SingleCellExperiment}}
#'input object. Gating is performed on cell-specific marker counts stored in the
#'\code{assay} slots of the object. These can either be raw counts (usually
#'stored in the \code{counts} slot) or transformed/scaled counts stored in other
#'assay slots. Gating can only be performed sample-wise; therefore, even if
#'\code{mask} or \code{image} are not specified, \code{img_id} needs to point to
#'the \code{colData} entry storing unique sample IDs. Furthermore, the
#'\code{cell_id} entry is required to identify cells during hierarchical gating.
#'
#'If mask is specified, marker expression values and selected cells will be
#'visualised by using the \code{\link{plotCells}} function. To visualise
#'pixel-level information with \code{\link{plotPixels}}, the user has to
#'further provide multi-channel images stored in a \code{CytoImageList}.
#'
#'@section The user interface (UI):
#'The UI's body is composed of two tabs (\emph{Scatter Plots} and
#'\emph{Images}). The side bar contains the \emph{General Control} and
#'\emph{Plots} panels. Both panels can be collapsed; however only one can be
#'expanded again.. Using the control panel, the user can set the number of
#'plots, which sample to display and which \code{assay} slot to use for plotting
#'expression. The \emph{Plots} panel can be used to select up to two markers per
#'plot. A one-marker selection is displayed as violin and jittered points, two
#'markers are shown as scatter plot. To define a subset of cells, the user can
#'draw gates on the plots. When using multiple plots, the gate applied in a plot
#'will define the subset of cells displayed in the next plot. The number of
#'plots, the marker selection and the applied gates will be transferred when
#'switching between samples. When switching markers or assay slots, the gates
#'will be removed.
#'
#'By switching to the \code{Images} tab, the user can inspect marker expression
#'levels and the result of the gating. This requires either an entry to
#'\code{mask} and (optionally) \code{image}. Depending on the input, the left
#'image visualises the expression of the selected markers by using
#'\code{\link{plotCells}} (if \code{mask} is supplied) or
#'\code{\link{plotPixels}} (if \code{image} is supplied). Displayed markers can
#'be changed using the dropdown menu. If \code{images} are provided, the user
#'can change the contrast of the selected markers. The right image colors
#'(\code{masks}) or outlines (\code{images}) gated cells. Both the left and the
#'right image come with a zoom-in functionality.
#'
#'The header section of the shiny allows to hide the side bar, to download
#'selected cells (see below), or to display the \code{sessionInfo} output and
#'the \emph{Help} section (see below).
#'
#'@section Download of gated cells:
#'The user can download the selected cells in form of a
#'\code{\linkS4class{SingleCellExperiment}} object. The output is a subset of
#'the input \code{SingleCellExperiment} object. The downloaded object (in form
#'of a .rds file) contains a new \code{colData} entry containing the label of
#'the subset defined by the user (using the \emph{Cell label} argument). The
#'cell label is stored in \code{colData(object)$cytomapper_CellLabel}
#'
#'The \code{metadata} slot of the \code{\linkS4class{SingleCellExperiment}}
#'object will be converted to a list. It contains the original metadata
#'(\code{metadata(object)$metadata}), gating parameters (e.g.
#'\code{metadata(object)$cytomapper_gate_1}), \code{sessionInfo()} output
#'(\code{metadata(object)$cytomapper_SessionInfo}), and the date
#'(\code{metadata(object)$cytomapper_GatingDate}).
#'
#'@section Passing further parameters:
#'To customise the visual output of \code{cytomapperShiny}, refer to
#'\code{\link{plotting-param}}. Further arguments can be passed to the
#'\code{\link{plotCells}} and \code{\link{plotPixels}} function. As an example:
#'To avoid interpolation of output images, set \code{interpolate = FALSE}.
#'Passing the parameter \code{interpolate = FALSE} will make images less blurry,
#'which might be useful while using the zoom-in functionality of
#'\code{cytomapperShiny} .
#'
#'@section Getting further help:
#'Please refer to the \emph{Help} section found when clicking the '?' button in
#'the upper right corner. The \emph{Help} section contains a recommended
#'workflow on how to use the app.
#'
#'@seealso \code{\link{plotCells}} and \code{\link{plotPixels}} for the main
#'plotting functions
#'
#'@return A Shiny app object for hierarchical gating of cells
#'
#'@examples
#'## Only run this example in interactive R sessions
#'if (interactive()) {
#'    # Load example data sets
#'    data("pancreasSCE")
#'    data("pancreasImages")
#'    data("pancreasMasks")
#'
#'    # Use shiny with SCE object, images and masks
#'    cytomapperShiny(object = pancreasSCE, mask = pancreasMasks, 
#'                  image = pancreasImages, cell_id = "CellNb", 
#'                  img_id = "ImageNb")
#'
#'    # Use shiny with SCE object and masks
#'    cytomapperShiny(object = pancreasSCE, mask = pancreasMasks,
#'                  cell_id = "CellNb", img_id = "ImageNb")
#'
#'    # Use shiny with SCE object only
#'    cytomapperShiny(object = pancreasSCE,
#'                  cell_id = "CellNb", img_id = "ImageNb")
#'}
#'
#'@author Nils Eling (\email{nils.eling@@dqbm.uzh.ch})
#'@author Tobias Hoch (\email{tobias.hoch@@dqbm.uzh.ch})
#'
#'@export
#'
#'@import shiny
#'@import shinydashboard
cytomapperShiny <- function(object,
                        mask = NULL,
                        image = NULL,
                        cell_id = NULL,
                        img_id = NULL,
                        ...) {
    # Object checks
    .valid.sce.shiny(object, img_id, cell_id, image, mask)

    if (!is.null(mask)) {
        .valid.sce(object, img_id, cell_id, exprs_values = NULL)
        .valid.mask(mask, img_id)
        .valid.matchObjects.plotCells(object, mask, img_id)
    }

    if (!is.null(image)) {
        if (is.null(mask)) {
            stop("Please provide a mask object.")
        }
        .valid.image(image, img_id)
        .valid.matchObjects.plotPixels(object, mask, image, img_id)
    }

    shiny_ui <- dashboardPage(
        header = .cytomapper_header(),
        sidebar = .cytomapper_sidebar(),
        body = .cytomapper_body(),
    )

    shiny_server <- function(input, output, session) {
        .cytomapper_server(object, mask, image, cell_id, img_id,
                            input, output, session, ...)
    }

    shinyApp(ui = shiny_ui, server = shiny_server)
}