R/sig_dist_matrix.R

In steverozen/ICAMSxtra: ICAMSxtra

#' Compute a matrix of distances / similarities between two sets of signatures.
#' 
#' @param x1 The first set of signatures (a positive matrix in which each column is a signature).
#'     The elements of \code{x1} will be the rows of the output matrix
#' 
#' @param x2 The second set of signatures, similar data type to \code{x1}.
#'     The elements of \code{x2} will be the columns of the output matrix
#'
#' @param method (as for the \code{philentropy::distance}) function.
#' 
#' @return A matrix with dimensions \code{ncol(x1)} X \code{ncol(x2)} with
#'   each element representing the distance or similarity (depending on \code{method})
#'   between the corresponding elements of \code{x1} and \code{x2}
#'
#' @export
#' 
#' @examples 
#' ex.sigs <- matrix(c(0.2, 0.8, 0.3, 0.7, 0.4, 0.6), nrow = 2)
#' colnames(ex.sigs) <- c("ex1", "ex2", "ex3")
#' gt.sigs <- matrix(c(0.21, 0.79, 0.19, 0.81), nrow = 2)
#' colnames(gt.sigs) <- c("gt1", "gt2")
#' sig_dist_matrix(ex.sigs, gt.sigs)

sig_dist_matrix <- function(x1, x2, method = "cosine") {
  mm <- cbind(x1, x2)
  dd <- suppressMessages(
    philentropy::distance(t(mm), method = method, use.row.names = TRUE))
  dd2 <- dd[1:ncol(x1), , drop = FALSE] # Use the rows that represent the elements of x1
  dd3 <- dd2[, -(1:ncol(x1)), drop = FALSE] # Use that columns that represent elements of x2
  return(dd3)
}