R/convert2hugo.R

In patterja/smmartprediction: SMMART Drug Prediction

#' convert ENSGids to HUGO gene names 
#' 
#' RNA matrix of ENSG identifiers converted and aggregated (summed) to HUGO
#' identifiers using targetid file from GENCODE v24. 
#' @param mat (matrix) columns are sample names, rows are ENSG IDs
#' @param targetid_file (char string) path to target_id.txt
#' @param protein_coding (logical): filter for protein coding only and then aggregate to HUGO
#' @return aggregated matrix with HUGO IDs
#' @export
#' @author Janice Patterson
#' @examples convert2hugo(mat)

convert2hugo <- function(mat, targetid_file="/Users/patterja/Box Sync/Heiser_SMMARTReport/REFERENCE_FILES/target_id.txt", combined=FALSE, protein_coding=FALSE){
  mat=as.matrix(mat)

  targetid = read.csv(targetid_file, sep="\t", stringsAsFactors = F)
  #filter by for protein coding
  if (protein_coding==TRUE){
    targetid <- targetid[which(targetid$X.7 == "protein_coding"),]}
 
  #remove redundant combinations of HUGO and ENSG
  targetid <- unique(targetid[, c("X.1", "X.5")])
  rownames(targetid) <- targetid$X.1

  #combined names
  targetid$combined1 =paste0(targetid$X.5, "_", targetid$X.1)
  targetid$combined2 =paste0(targetid$X.5, "_", gsub("\\.[0-9]+$","",targetid$X.1))
  
  if (combined==TRUE){
    rownames(mat) = targetid$combined2[match(rownames(mat), (targetid$X.1))]
  } else{
    #filter the mat for the targets
    mat = mat[intersect(targetid[,"X.1"], rownames(mat)),]
    #convert to HUGO
    rownames(mat) = targetid$X.5[match(rownames(mat), targetid$X.1)]
  }
  #redundant HUGO names are summed
  mat_gene = aggregate(mat,by=list(rownames(mat)), FUN=sum)
  row.names(mat_gene) <- mat_gene[[1]]
  mat_gene[[1]]=NULL
  return(mat_gene)
}