R/internal_rm_duplicated_biopax_pw.R

In grishagin/RIGbiopax: Extract and Compare Genes and Pathways from BioPAX

internal_rm_duplicated_biopax_pw<-
  function(big_df){
    
    #' @keywords internal
    
    library(plyr)
    library(dplyr)
    
    #add gene match indicators
    .genematch<-
      rep(NA,nrow(big_df))
    #match
    .genematch[which(big_df$ENTREZID==big_df$toxdb.Gene.ID)]<-
      "1_match"
    #only in toxdb list
    .genematch[which(is.na(big_df$biopax.Gene.ID.Type) & !is.na(big_df$toxdb.Gene.ID))]<-
      "2_toxdb"
    #only in biopax
    .genematch[which(!is.na(big_df$biopax.Gene.ID.Type) & is.na(big_df$toxdb.Gene.ID))]<-
      "3_biopax"
    #neither in biopax nor in the toxdb list
    .genematch[which(is.na(big_df$biopax.Gene.ID.Type) & is.na(big_df$toxdb.Gene.ID))]<-
      "4_genes_not_found"
    #add column to the big df with genes
    big_df <-
      big_df%>%
      mutate(.genematch=.genematch) 
    
    #filter out individual toxdb/biopax id pairs
    #and match counts
    comp_df<-
      big_df %>%
      filter(!is.na(toxdb.Pathway.ID)) %>%
      group_by(toxdb.Pathway.ID
               ,biopax.Pathway.ID
               ,.genematch) %>%
      dplyr::summarise(count=n()) %>%
      spread(key=.genematch
             ,value=count
             ,fill = 0) %>%
      mutate(sum=`1_match`+`2_toxdb`+`3_biopax`)
    
    dupl_txid<-
      comp_df$toxdb.Pathway.ID[duplicated(comp_df$toxdb.Pathway.ID)]
    
    #for each duplicate txid, find biopax ids to be removed
    #because there has to be only one!
    bad_bpids<-
      dupl_txid %>%
      lapply(function(txid){
        #filter by tox id
        #and arrange by max match, then max sum
        #take all but the first bp ids
        bad_bpid<-
          comp_df %>%
          filter(toxdb.Pathway.ID %in% txid) %>%
          dplyr::arrange(desc(`1_match`)
                         ,desc(sum)
          ) %>%
          .$biopax.Pathway.ID %>%
          .[-1]
        return(bad_bpid)
      }) %>%
      unlist
    
    #clean big df
    big_df<-
      big_df %>%
      filter(!biopax.Pathway.ID %in% bad_bpids)
    
    return(big_df)
  }