R/normalize_exprs_mat.R

In tinyheero/RHL30: An R package for the RHL30 Predictor

#' Generate the Normalizer Values for each Sample
#'
#' `get_sample_normalizer_value` returns the normalizer values for each 
#' sample. The normalizer value is based on the geometric mean of the input
#' `normalizer_genes`
#'
#' @param exprs_mat The expression matrix.
#' @param normalizer_genes Character vector containing the genes to use for 
#'   calculating the normalizer
#' @return A numeric vector with the samples as names and the normalizer 
#'   as values
#' @export
get_sample_normalizer_value <- function(exprs_mat, normalizer_genes) {

  message(
    "[get_normalizer]: Generating the geometric mean of housekeeper genes"
  )

  if (! length(normalizer_genes) > 1) {
    stop("[get_normalizer]: Must have more than 1 housekeeper gene")
  }

  if (! all(normalizer_genes %in% rownames(exprs_mat))) {
    missing_ind <- !normalizer_genes %in% rownames(exprs_mat)
    stop(
      paste0(
        "[get_normalizer]: Matrix is missing the genes: ", 
        normalizer_genes[missing_ind])
    )
  }

  normalizer_log <- log(exprs_mat[normalizer_genes, ])
  normalizer_log_mean <- apply(normalizer_log, 2, mean)
  normalizer_log_mean_exp <- exp(normalizer_log_mean)
  return(normalizer_log_mean_exp)
}

#' Normalize the Expression Matrix 
#'
#' This function normalizes the expression matrix.
#' 
#' @param exprs_mat The expression matrix
#' @param sample_normalizer_values Sample normalizer values generated by 
#'   the `get_sample_normalizer_value` function
#' @return A normalized expression matrix
#' @export
normalize_exprs_mat <- function(exprs_mat, sample_normalizer_values) {

  samples <- colnames(exprs_mat)
  if (! all(samples %in% names(sample_normalizer_values) )) {
    stop(
      paste(
        "[normalize_exprs_mat]: Not all samples in the exprs_mat have sample",
        "normalizer values"
      )
    )
  }
  sample_normalizer_values <- sample_normalizer_values[samples]

  message("[normalize_exprs_mat]: Normalizing the expression exprs_matrix")
  exprs_mat_norm <- sweep(exprs_mat, 2, sample_normalizer_values, "/")
  exprs_mat_norm_scaled <- exprs_mat_norm * 1000

  # Ensure that any values < 1 do not become negative
  exprs_mat_norm_scaled[exprs_mat_norm_scaled < 1] <- 1

  message("[normalize_exprs_mat]: Log2 transforming")
  exprs_mat_norm_scaled_log <- log(exprs_mat_norm_scaled, 2)
  return(exprs_mat_norm_scaled_log)
}