annotate clean_rm_output.R @ 2:7f1032da7a0a draft

Uploaded

#!/usr/bin/env Rscript
suppressPackageStartupMessages(library(rtracklayer))
suppressPackageStartupMessages(library(parallel))


gff_cleanup = function(gff){
  ## remove overlapin annotation track - assign new annot
  gff_disjoin = disjoin(gff, with.revmap=TRUE)
  ## append annotation:
  gff_names = mclapply(as.list(gff_disjoin$revmap), FUN = function(x)gff$Name[x], mc.cores = 8)
  gff_strands = mclapply(as.list(gff_disjoin$revmap), FUN = function(x)strand(gff[x]), mc.cores = 8)
  new_annot = sapply(sapply(gff_names, unique), paste, collapse="|")
  new_annot_uniq = unique(new_annot)
  lca_annot = sapply(strsplit(new_annot_uniq, "|", fixed = TRUE), resolve_name)
  names(lca_annot) = new_annot_uniq
  new_annot_lca = lca_annot[new_annot] 
  #new_annot_lca = sapply(sapply(gff_names, unique), resolve_name)
  strand_attribute = sapply(sapply(gff_strands, unique), paste, collapse="|")
  gff_disjoin$source="RM"
  gff_disjoin$type="repeat"
  gff_disjoin$score=NA
  gff_disjoin$phase=NA
  gff_disjoin$Name=new_annot_lca
  gff_disjoin$Original_names=new_annot
  gff_disjoin$strands=strand_attribute
  gff_disjoin$revmap=NULL
  return(gff_disjoin)
}

resolve_name=function(x){
  if (length(x)==1){
    # no conflict
    return(x)
  } else{
    y = sapply(x, strsplit,  split="/", fixed = TRUE)
    ny = table(unlist(sapply(y, function(x)paste(seq_along(x),x))))
    if (max(ny)<length(x)){
      return("Unknown")
    }else{
      k = which(ny==length(x))
      r = max(as.numeric((gsub(" .+","",names(k)))))
      out = paste(y[[1]][1:r], collapse="/")
      return(out)
    }
  }
}

convert_names <- function(n, old_sep = "|" , new_sep = "\""){
  # remove all characters which are new_sep with -
  n_new = gsub(old_sep, new_sep,
               gsub(new_sep,"-", n, fixed = TRUE),
               fixed = TRUE)
  return(n_new)
}


infile = commandArgs(T)[1]
outfile = commandArgs(T)[2]


## infile = "./test_data/raw_rm.out"

rm_out = read.table(infile, as.is=TRUE, sep="", skip = 2, fill=TRUE, header=FALSE, col.names=paste0("V",1:16))

gff = GRanges(seqnames = rm_out$V5, ranges = IRanges(start = rm_out$V6, end=rm_out$V7))

# repeat class after # symbol - syntax 1
# detect separator
# if "|" is present replace "|" -> "/" and "/" -> "-"
if (any(grepl("|", rm_out$V11, fixed = TRUE))){
  gff$Name <- convert_names(rm_out$V11, old_sep = "|", new_sep = "/")
  message('replacing classification separator character "|" with "/"')
  print(gff)
}else{
  gff$Name <- rm_out$V11
}

## is repeat type is specifies by double underscore:
## then rm_out$V11 is unspecified
if (any(rm_out$V11 == "Unspecified")){
  ## set Name from prefix
  ## TODO
  inc = rm_out$V11 == "Unspecified"
  Name = gsub("__.+","",rm_out$V10)
  # chanche Usnpsecified to new name
  gff$Name[inc] = Name
}


## join neighbors with the same annotation, disregard strand!
result <- unlist(reduce(split(gff, gff$Name)))

result$Name <- names(result)

result_clean = gff_cleanup(result)

## TODO
## identify conflicting annotation, replace by LCA but keep origin list of classifications

gff_out = sortSeqlevels(result_clean)
gff_out = sort(gff_out)
gff_out$type = "repeat_region"
gff_out$source = "RepeatMasker_parsed"
gff_out$ID=paste0(gff_out$Name, "_", seq_along(gff_out$Name))
export(gff_out,  format = "gff3", con=outfile)