TCGAanalyze_RemoveReplicateSamples.R

# http://gdac.broadinstitute.org/runs/sampleReports/latest/READ_Replicate_Samples.html
TCGAanalyze_RemoveReplicateSamples<-function(tabDF, typesample){
  tabDF.typesample <- tabDF[, TCGAquery_SampleTypes(colnames(tabDF), typesample=typesample)]
  # detect duplicated patient for typesample
  tabDF.typesample.duplicated = duplicated(tabDF.typesample$patient)
  if (sum(tabDF.typesample.duplicated)>0){
    tabDF.typesample.not_duplicated = !(tabDF.typesample$patient %in% tabDF.typesample[,tabDF.typesample.duplicated]$patient)
    # print(tabDF.typesample$patient[tabDF.typesample.duplicated])
    # print(class(tabDF.typesample$patient[tabDF.typesample.duplicated]))
    barcodes_solved_duplication=sapply(tabDF.typesample$patient[tabDF.typesample.duplicated],function(patient, dataset){
      barcodes = dataset[,dataset$patient==patient]$barcode
      # search for H
      
      H_barcodes = sapply(barcodes, function(barcode){
        stringi::stri_sub(stringi::stri_split(str = barcode, regex = "-")[[1]][5], -1) == "H"
      })
      
      if (sum(H_barcodes)==0){
        R_barcodes = sapply(barcodes, function(barcode){
          stringi::stri_sub(stringi::stri_split(str = barcode, regex = "-")[[1]][5], -1) == "R"
        })
        if (sum(R_barcodes)==0){
          warning("Removing duplicates, there is no H or R in the duplicated barcodes.")
        } else {
          barcodes = barcodes[R_barcodes]
        }
      } else {
        barcodes = barcodes[H_barcodes]
      }
      
            barcodes = sort(x=barcodes,decreasing = TRUE)
      print(barcodes[1])
      return(barcodes[1])
    }, tabDF.typesample)
    
    
    return(tabDF.typesample[,c(tabDF.typesample[,tabDF.typesample.not_duplicated]$barcode, barcodes_solved_duplication)])
  } else {
    return(tabDF.typesample)
  }
}