genome_prepare/script/ensembl_gbk2tsv.pl

#! /usr/bin/perl -w
#This file is designed to extract annotation information from ensembl\plantgbd genebank file.
#The file has 9 colume define as:
#Gene ID;      Start Site;      End Site;       Direction;      Locus tag;      GO terms;      EntrezGene ID;      NR Description
#perl <genebankfile> 

use strict;
use warnings;
my $usage="USAGE: perl $0 <DAT.gz> <gbk>\n\n";

die $usage if(@ARGV==0);


#open GENEBANK, "<$ARGV[0]" or die $!;
if($ARGV[0]=~/gz$/){
	open GENEBANK, "gunzip -dc $ARGV[0]|" or die $!;
}else{
	open GENEBANK, "$ARGV[0]" or die $!;
}
my ($Chr, $Start, $End, @Gene, $Ingene, $note_sign, $id, $i, $j);

$Ingene = 0;
$note_sign = 0;
$id = -1;
my $n;
while(<GENEBANK>){
	#print $n++,"_", $Ingene,"_", $note_sign,"_", $id, $_;
	chomp;
	if(/^LOCUS/){($Chr) = /^LOCUS       (\S+) .*/} 
	if(/^..\s{3,5}\S+\s+/){$Ingene=0};
	if(/^..\s{3,5}gene\s+/){
		$id ++;
		if(($Start, $End) = /complement\(([1-9][0-9]*)\.\.([1-9][0-9]+)\)/){
			$Gene[$id][4] = "-";
			$Gene[$id][2] = $1;
			$Gene[$id][3] = $2;
			$Ingene = 1;
			next;
			}
		elsif(($Start, $End) = /([1-9][0-9]*)\.\.([1-9][0-9]+)/){
			$Gene[$id][4] = "+";
			$Gene[$id][2] = $1;
			$Gene[$id][3] = $2;
			$Ingene = 1;
			next;
			}
		else{
			die "Error encounter when read gene lenghth."
			}
	}
	if($Ingene){
		if($_ =~ /\/gene=([^"]+)$/){$Gene[$id][0] = $1;$Gene[$id][1] = $Chr;next;}
		elsif($_ =~ /\/locus_tag="(.+)"$/){$Gene[$id][5] = $1;next;}
		elsif($_ =~ /\/note="(.+)(")?$/){$Gene[$id][8] = $1; if($2){$note_sign = 0;}else{$note_sign = 1;}next;}
		else{if($note_sign && $_ =~ /^..\s{7,}([^"]+)(")?$/){$Gene[$id][8] .= $1;next;}else{if($note_sign == 1){$Ingene = 0;}$note_sign = 0;}}
	}
	else{
		if(/db_xref=".*(GO:[\d]{7,7})"/){$Gene[$id][6] .= $1.";"}
		if(/db_xref="(EntrezGene:[\d]{1,})"/){$Gene[$id][7] .= $1.";"}}
	
}
open R,">$ARGV[1]";
for $i (0 .. $#Gene ){
	for $j (0 .. 8){
		defined($Gene[$i][$j]) or $Gene[$i][$j] ="-";
		print R "$Gene[$i][$j]\t";
		}
	print R "\n";
	} 

close GENEBANK;
close R;