R/feature_filter.R

In tpall/featurefilter: Filter out non-informative HT-seq read counts

#' Filter out features with low counts
#'
#' Uses sequence read counts and number of samples per group to filter
#' out genes with low number of read counts in most samples.
#'
#' @param rawcounts counts matrix
#' @param n_samples minimum number of samples passing the filter
#'
#' @return logical vector
#'
#' @importFrom edgeR cpm
#'
#' @export
#'
feature_filter <- function(rawcounts, n_samples) {

  # Calculate normalised counts
  cpms <- edgeR::cpm(rawcounts)

  # Smallest libsize in millions of counts
  libsize <- min(colSums(rawcounts, na.rm = TRUE)) / 1e6

  # Library size adjusted feature filter
  rowSums(cpms > 10 / libsize) >= n_samples
}