Efficient calculation of percentage of expression per cluster, for example for scRNA-seq.

getPercentExpression.R

#' Given a matrix-like object with genes being rows, samples/cells being columns,
#' and a group information for the columns, calculate the percentage of expression per group. 
#' No dependencies other than base R. The CsparseMatrix format, commonly used in scRNA-seq is supported.
#'
#' @param data numeric matrix or data.frame-like object
#' @param group a vector (character or factor)
#' @param threshold values above this threshold are considered expressed
#' @param digits round results to this number of digits
#'
#' @examples
#' data <- matrix(rnorm(100, 2, 2), ncol=10)
#' group <- rep(c(1,2), each=5)
#' getPercentExpression(data, group, 0, 2)
#' 
getPercentExpression <- function(data, group, threshold=0, digits=2){
  
  if(ncol(data)!=length(group)) stop("ncol(data) != length(group)")
  if(!is.numeric(threshold) | threshold < 0) stop("threshold must be numeric and > 0")
  
  datar <- (data > threshold) * 1
  a <- base::rowsum(x=as.matrix(t(datar)), group=group)
  b <- as.numeric(table(group)[rownames(a)])
  f <- round(100*t(a/b), digits=digits)
  f
  
}