R/plotSuggestedK.R

In navinlabcode/copykit: CopyKit

#' plotSuggestedK
#'
#' Uses the information from \code{\link{findSuggestedK}} to plot the values
#' of jaccard similarity from the tested k range on \code{\link{findSuggestedK}}.
#'
#' @param scCNA The scCNA object.
#' @param geom A character with the geom to be used for plotting.
#'
#' @details \code{\link{plotSuggestedK}} access the \code{\link[S4Vectors]{metadata}}
#' element suggestedK_df that is saved to the scDNA object after running
#' \code{\link{findSuggestedK}}. The dataframe is used for plotting either a
#' heatmap, when the argument geom = 'tile', or a dotplot when argument geom =
#' 'dotplot' or a boxplot when geom = 'boxplot'.
#'
#' \itemize{
#' \item{geom = 'boxplot':} Plots a boxplot of the jaccard similarities across
#' all clusters detected in the grid search. The large red points represent
#' the mean jaccard similarity.
#'
#' #' \item{geom = 'tile':} Plots a heatmap of the jaccard similarities across
#' all clusters detected in the grid search. The filling collors represent the
#' jaccard similarity value. Rows represent clusters and columns the k value.
#'
#' #' \item{geom = 'dotplot':} Plots a dotplot of the jaccard similarities across
#' all clusters detected in the grid search. Where the size of the dots represent
#' the jaccard similarity for each assesed k value.
#'
#' #' \item{geom = 'scatterplot':} Plots a scatterplot of the jaccard similarity
#' explained by the number of cells. Points are colored by subclone and lines
#' represent a linear regression across the points.
#'
#' }
#'
#' @return A ggplot2 object with the plot of different tested k values and their
#' jaccard similarity for each subclone
#'
#' @export
#'
#' @import ggplot2
#' @importFrom dplyr mutate
#' @importFrom tidyr complete
#' @importFrom S4Vectors metadata
#' @importFrom gtools mixedsort
#'
#' @examples
#' copykit_obj <- copykit_example_filtered()
#' copykit_obj <- findSuggestedK(copykit_obj)
#' plotSuggestedK(copykit_obj)
plotSuggestedK <- function(scCNA,
                           geom = c("boxplot", "tile", "dotplot", "scatterplot")) {
    geom <- match.arg(geom)

    # bindings for NSE objects
    k <- subclones <- bootmean <- chosen <- mean_jac <- n_cells <- NULL

    df <- S4Vectors::metadata(scCNA)$suggestedK_df
    sug_k <- S4Vectors::metadata(scCNA)$suggestedK

    df <- dplyr::mutate(df, k = as.character(k))

    # df expanded for geom tile
    df_exp <- tidyr::complete(
        df,
        k,
        subclones
    )

    # common layers
    common_layers <- list(
        scale_y_discrete(limits = gtools::mixedsort(unique(df$subclones))),
        scale_x_discrete(limits = gtools::mixedsort(unique(df$k))),
        theme_classic(), labs(fill = "jaccard\nsimilarity")
    )

    if (geom == "dotplot") {
        p <- ggplot(df, aes(k, subclones)) +
            geom_point(aes(
                size = bootmean,
                fill = bootmean
            ),
            shape = 21
            ) +
            scale_fill_viridis_c(option = 2) +
            common_layers
    }

    if (geom == "tile") {
        p <- ggplot(df_exp, aes(k, subclones)) +
            geom_tile(aes(fill = bootmean), color = "black") +
            scale_fill_viridis_c(na.value = "grey", option = 2) +
            common_layers +
            theme(panel.border = element_rect(fill = NA, size = 3))
    }

    if (geom == "boxplot") {
        mean_per_k <- df %>%
            dplyr::group_by(k) %>%
            dplyr::summarise(mean_jac = mean(bootmean))

        # adding color for chosen k
        df <- dplyr::mutate(df, chosen = ifelse(df$k == as.character(sug_k),
            TRUE,
            FALSE
        ))

        p <- ggplot() +
            geom_boxplot(
                data = df, aes(k, bootmean, fill = chosen),
                alpha = 1
            ) +
            geom_point(
                data = mean_per_k, aes(k, mean_jac),
                fill = "red",
                shape = 21,
                size = 3
            ) +
            scale_fill_manual(values = c("TRUE" = "khaki", "FALSE" = "grey90")) +
            scale_x_discrete(limits = gtools::mixedsort(unique(df$k))) +
            theme_classic() +
            theme(legend.position = "none",
                  axis.title.x = element_text(size = 14),
                  axis.text.x = element_text(size = 12),
                  axis.title.y = element_text(size = 14),
                  axis.text.y = element_text(size = 12)) +
            labs(y = "jaccard similarity")
    }

    if (geom == "scatterplot") {
        p <- ggplot(df, aes(x = n_cells, y = bootmean)) +
            geom_point(aes(fill = subclones), shape = 21) +
            stat_smooth(method = "lm", se = FALSE) +
            facet_wrap(vars(as.numeric(k)), scales = "free_x") +
            scale_fill_manual(
                values = subclones_pal(),
                limits = gtools::mixedsort(unique(df$subclones))
            ) +
            theme_classic() +
            labs(
                x = "number of cells",
                y = "jaccard similarity"
            )
    }

    # return plot
    p
}