R/data_preprocess.R

In scSorter: Implementation of 'scSorter' Algorithm

#' Preprocess Data
#'
#' This function validates and preprocesses the input data for the downstream analysis.
#' @keywords internal
#'
#' @param expr A matrix of input data. Each row represents a gene and each column represents a cell.
#' @param anno_processed A list of processed annotation information that consists of the design matrix and the weight matrix for marker genes.
#'
#' @return A list contains processed expression matrix, design matrix, and weight matrix.
#'

data_preprocess = function(expr, anno_processed) {
  designmat = anno_processed[[1]]
  weightmat = anno_processed[[2]]

  rownames(expr) = toupper(rownames(expr))
  rownames(designmat) = toupper(rownames(designmat))

  markers = rownames(designmat)

  markers_avail = markers %in% rownames(expr)

  if (sum(markers_avail) < length(markers)) {
    warning(paste('The following specified marker genes are not found from the expression data: ',
                  paste(markers[!markers_avail], collapse = ', '), '.', sep = ''))

    designmat = designmat[markers_avail, ]
    csdmat = colSums(designmat)

    if(sum(csdmat > 0) < ncol(designmat)) {
      stop(paste('None of the marker genes specfied for the following cell types could be found from the expression data: ',
                    paste(colnames(designmat)[csdmat == 0], collapse = ', '), '. \n Please maker sure the marker gene names are correctly specified.', sep = ''))
    }
  }

  dmat_gene_names = rownames(designmat)
  dat_gene_names = rownames(expr)

  picker = dat_gene_names %in% dmat_gene_names

  expr_mk = expr[picker, ]
  expr_rt = expr[!picker,]

  #reorder genes so that the order of expr mat and design mat matches
  ror = rep(0, nrow(designmat))
  for(i in 1:nrow(designmat)){
    ror[i] = which(rownames(expr_mk) == dmat_gene_names[i])
  }
  expr_mk = expr_mk[ror,]

  expr_cb = rbind(expr_mk, expr_rt)

  return(list(dat = expr_cb, designmat = designmat, weightmat = weightmat))
}