R/calculate_entropy.R

In TrigosTeam/SPIAT: Spatial Image Analysis of Tissues

#' calculate_entropy
#' 
#' @description If arg `radius` is not specified, the function returns the
#'   entropy of the cell types of interest for the whole image. If arg `radius`
#'   is specified, the function returns a data frame where each row is a
#'   reference cell and the columns stores the entropy of the cell types of
#'   interest in each circle of the reference cells.
#' @param spe_object SpatialExperiment object in the form of the output of
#'   \code{\link{format_image_to_spe}}.
#' @param cell_types_of_interest String Vector. Cell types of interest. If arg
#'   `radius` is not NULL, the first cell type is considered as reference cell
#'   type. Circles of the specified radius will be drawn around the reference
#'   cells and the entropy of cell types will be calculated for each of the
#'   reference cells.
#' @param feature_colname String specifying the column the cell types are from.
#' @param radius (OPTIONAL) Numeric. The maximum radius around a reference cell
#'   for another cell to be considered an interaction.
#' @return A dataframe or a number depending on the argument radius
#' @export
#' @examples
#' calculate_entropy(SPIAT::defined_image, 
#' cell_types_of_interest = c("Immune1","Immune2"),
#' feature_colname = "Cell.Type")

calculate_entropy <- function(spe_object, cell_types_of_interest, 
                              feature_colname = "Phenotype", radius = NULL){
    
    if((cell_types_of_interest[1] %in% 
        SummarizedExperiment::colData(spe_object)[,feature_colname]) && 
       any(cell_types_of_interest[-1] %in% 
           SummarizedExperiment::colData(spe_object)[,feature_colname])){
        if (!is.null(radius)){
            reference_marker <- cell_types_of_interest[1]
            target_marker<-cell_types_of_interest
            n_cells <- number_of_cells_within_radius(
                spe_object, reference_celltype = reference_marker,
                target_celltype = target_marker,radius = radius,
                feature_colname = feature_colname)
            n_cells.df <- n_cells[[reference_marker]]
            n_cells.df[,"total"] <- 0
            
            for (i in target_marker){
                n_cells.df[,paste(i,"_log2",sep = "")] <- log2(n_cells.df[,i])
                n_cells.df[,"total"] <- n_cells.df[,"total"]+n_cells.df[,i]}
            n_cells.df[,"total_log2"] <- log2(n_cells.df[,"total"])
            n_cells.df[n_cells.df == -Inf] <- 0
            
            for (i in target_marker){
                n_cells.df[which(n_cells.df[,"total"] != 0), 
                           paste(i,"ratio",sep = "")] <- 
                    n_cells.df[,i]/n_cells.df[,"total"]
                n_cells.df[which(n_cells.df[,"total"] == 0), 
                           paste(i,"ratio",sep = "")] <- 0
                n_cells.df[,paste(i,"_entropy",sep = "")] <- 
                    n_cells.df[, paste(i,"ratio",sep = "")] *
                    (n_cells.df[,paste(i,"_log2",sep = "")] - 
                         n_cells.df[,"total_log2"])}
            n_cells.df[,"entropy"] <- 
                -(rowSums(n_cells.df[,grepl("_entropy",colnames(n_cells.df))]))  
            
            return(n_cells.df)} else{
            entropy_all <- 0
            data <- data.frame(SummarizedExperiment::colData(spe_object))
            data <- data[which(data[,feature_colname] %in% 
                                   cell_types_of_interest),]
            n_all <- dim(data)[1]
            
            if (n_all == 0){
                return(0.0)} else{
                for (type in cell_types_of_interest){
                    type_data <- data[which(data[,feature_colname]==type),]
                    n_type <- dim(type_data)[1]
                    if (n_type != 0){
                        p_type <- n_type/n_all
                        entropy_type <- -(p_type)*log2(p_type)
                        entropy_all <- entropy_all + entropy_type}}
                return(entropy_all)}}}else{
        methods::show("Cell type not found!")
        return(NA)}}