#' plot_marker_level_heatmap
#'
#' @description Blurs the image by splitting the images into small squares. The
#' marker levels are then averaged within each square. All cells are
#' considered, regardless of phenotype status.
#'
#' @param spe_object SpatialExperiment object in the form of the output of
#' \code{\link{format_image_to_spe}}.
#' @param marker String. Marker to plot.
#' @param num_splits Integer specifying the blurring level (number of splits)
#' for the image. Higher numbers result in higher resolution.
#' @import dplyr
#' @import ggplot2
#' @return A plot is returned
#' @examples
#' plot_marker_level_heatmap(SPIAT::simulated_image, num_splits = 100, "Tumour_marker")
#' @export
plot_marker_level_heatmap <- function(spe_object, num_splits, marker){
# setting these variables to NULL as otherwise get "no visible binding for global variable" in R check
xcord <- ycord <- NULL
formatted_data <- get_colData(spe_object)
intensity_matrix <- SummarizedExperiment::assay(spe_object)
markers <- rownames(intensity_matrix)
#CHECK
if (is.element(marker, markers) == FALSE) {
stop("The marker specified is not in the data")
}
cell_ids <- colnames(intensity_matrix)
rownames(intensity_matrix) <- NULL
colnames(intensity_matrix) <- NULL
intensity_matrix_t <- t(intensity_matrix)
intensity_df <- data.frame(intensity_matrix_t)
colnames(intensity_df) <- markers
formatted_data <- cbind(formatted_data, intensity_df)
formatted_data <- formatted_data[stats::complete.cases(formatted_data),]
formatted_data$split.X <- 0
formatted_data$split.Y <- 0
minX <- min(formatted_data$Cell.X.Position, na.rm = TRUE)
maxX <- max(formatted_data$Cell.X.Position, na.rm = TRUE)
minY <- min(formatted_data$Cell.Y.Position, na.rm = TRUE)
maxY <- max(formatted_data$Cell.Y.Position, na.rm = TRUE)
#Splits the range of x and y coordinates
#into n + 1 evenly spaced out lengths
x_split <- seq(minX, maxX, length.out = num_splits + 1)
y_split <- seq(minY, maxY, length.out = num_splits + 1)
#Creates matrix of the locations of x and y cuts to the image
split_occurrence <- cbind(x_split, y_split)
#obtain the x and y coordinates on a heatmap for every cell based on number of splits
for (y in seq_len(num_splits)){
local_coor_y <- y_split[c(y+1, y)]
#grab the cells in the range
result <- formatted_data[min(local_coor_y) < formatted_data$Cell.Y.Position & formatted_data$Cell.Y.Position <= max(local_coor_y), ]
if(y == 1){
extra_row <- formatted_data[formatted_data$Cell.Y.Position == min(local_coor_y), ]
result <- rbind(result, extra_row)
}
if(nrow(result) > 0) {
result$split.Y <- y
formatted_data[match(result$Cell.ID,formatted_data$Cell.ID),] <- result
}
}
for (x in seq_len(num_splits)){
local_coor_x <- x_split[c(x+1, x)]
#grab the cells in the range
result <- formatted_data[min(local_coor_x) < formatted_data$Cell.X.Position & formatted_data$Cell.X.Position <= max(local_coor_x), ]
if(x == 1){
extra_row <- formatted_data[formatted_data$Cell.X.Position == min(local_coor_x), ]
result <- rbind(result, extra_row)
}
if(nrow(result) > 0) {
result$split.X <- x
formatted_data[match(result$Cell.ID,formatted_data$Cell.ID),] <- result
}
}
heatmap_title <- paste(marker, "level")
#create a df with only the intensity level of a single marker of interest and the coordinates
df <- stats::aggregate(formatted_data[,marker], by=list(xcord=formatted_data$split.X, ycord=formatted_data$split.Y), FUN=mean)
p <- ggplot(df, aes(xcord, ycord, fill=x)) + geom_tile()
p <- p + scale_fill_gradient(low="white", high="red")
p <- p + xlab("x position") + ylab("y position")
p <- p + labs(fill = "Mean intensity level") + ggtitle(heatmap_title)
p <- p + theme(panel.background = element_rect(fill = "grey", colour = "grey", linetype = "solid"),
panel.grid.major = element_blank(),
panel.grid.minor = element_blank())
# methods::show(p)
return(p)
}
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.