| 1 |
#!/usr/bin/perl |
| 2 |
use strict; |
| 3 |
use Getopt::Std; |
| 4 |
use LWP::Simple; |
| 5 |
use FindBin;use lib $FindBin::Bin; |
| 6 |
use dbSession; |
| 7 |
my $url= |
| 8 |
'http://www.geneontology.org/ontology/gene_ontology_edit.obo'; |
| 9 |
my $usage = qq/Usage: |
| 10 |
go2bcp.pl [-f <input_obo_file>] |
| 11 |
|
| 12 |
Parses and loads the GO data in geanno db. |
| 13 |
The GO data are either downloaded from the embedded url: |
| 14 |
$url |
| 15 |
..or, if -f option is given, from a local .obo file downloaded previously. |
| 16 |
/; |
| 17 |
umask 0002; |
| 18 |
getopts('f:') || die($usage."\n"); |
| 19 |
my $file=$Getopt::Std::opt_f; |
| 20 |
unless ($file) { |
| 21 |
print STDERR "downloading file..\n"; |
| 22 |
$file='gene_ontology_data.obo'; |
| 23 |
unlink($file); |
| 24 |
die("Error: file $file already exists (couldn't remove!)\n") |
| 25 |
if -f $file; |
| 26 |
die "Error downloading with getstore()!\n" |
| 27 |
if is_error(getstore($url, $file)); |
| 28 |
die "Error: file $file is non-existent or zero size after retrieval!\n" |
| 29 |
unless -s $file; |
| 30 |
print STDERR "Download OK.\n"; |
| 31 |
} |
| 32 |
|
| 33 |
my %relcode=( 'part_of' => 'P', |
| 34 |
'is_a' => 'I', |
| 35 |
'alt_id'=> 'A', |
| 36 |
'upd_ann'=>'U' |
| 37 |
); |
| 38 |
my %namespace=( |
| 39 |
'biological_process'=>'P', |
| 40 |
'cellular_component'=>'C', |
| 41 |
'molecular_function'=>'F' |
| 42 |
); |
| 43 |
#create the bcp files for loading |
| 44 |
|
| 45 |
open(GOTERM, '>go_term.bcp') || die ("Error creating file go_term.bcp!\n"); |
| 46 |
open(GOLINK, '>go_link.bcp') || die ("Error creating file go_link.bcp!\n"); |
| 47 |
open(GOSYN, '>go_synonyms.bcp') || die ("Error creating file go_synonyms.bcp!\n"); |
| 48 |
open(GOXREF, '>go_xref.bcp') || die("Error creating file go_xref.bcp!\n"); |
| 49 |
open(GOSET, '>go_subset.bcp') || die("Error creating file go_subset.bcp!\n"); |
| 50 |
|
| 51 |
#my $ds=dbSession->new('geanno@NEOSYBASE'); |
| 52 |
|
| 53 |
open(FGO, $file) || die ("Error opening $file\n"); |
| 54 |
|
| 55 |
my $c_id; #current go term being parsed |
| 56 |
my $c_type; # 'P', 'F', or 'C' |
| 57 |
my $c_name; |
| 58 |
my $c_def; |
| 59 |
my $c_obsolete=0; |
| 60 |
my $c_comment; |
| 61 |
while (<FGO>) { |
| 62 |
if (m/^\s*\[Term\]/) { |
| 63 |
goTerm() if $c_id; |
| 64 |
my $id; |
| 65 |
do { $_=<FGO>; |
| 66 |
($id)=(/^id:\s*GO\:(\d+)/); |
| 67 |
} until $id || !$_; |
| 68 |
die("Couldn't parse current term ID!\n") unless $id; |
| 69 |
$c_id=int($id); |
| 70 |
($c_name, $c_type, $c_def, $c_comment)=(undef,undef,undef,undef); |
| 71 |
$c_obsolete=0; |
| 72 |
next; |
| 73 |
} |
| 74 |
next unless $c_id; |
| 75 |
next if m/^\s*$/; |
| 76 |
chomp; |
| 77 |
s/\s+$//;tr/\t/ /s; |
| 78 |
if (/^name:\s*(.+)/) { |
| 79 |
$c_name=$1; |
| 80 |
} |
| 81 |
elsif (/^namespace:\s*(\S+)/) { |
| 82 |
$c_type=$namespace{lc($1)}; |
| 83 |
} |
| 84 |
elsif (/^def:\s*(.+)/) { |
| 85 |
$c_def=$1; |
| 86 |
$c_def =~ tr/"//d; #"#-- |
| 87 |
} |
| 88 |
elsif (/^is_obsolete:\s*true/) { |
| 89 |
$c_obsolete=1; |
| 90 |
} |
| 91 |
elsif (/^relationship:\s*(\S+)\s*GO:(\d+)/) { |
| 92 |
my ($rel, $id2)=($1, int($2)); |
| 93 |
$rel=$relcode{lc($rel)} || |
| 94 |
die("Error: unrecognized relationship code '$rel' for term $c_id ($_)\n"); |
| 95 |
print GOLINK join("\t",$c_id, $rel, $id2)."\n"; |
| 96 |
} |
| 97 |
elsif (/^alt_id:\s*GO:(\d+)/) { |
| 98 |
my $id2=int($1); |
| 99 |
print GOLINK join("\t",$c_id, $relcode{'alt_id'}, $id2)."\n"; |
| 100 |
} |
| 101 |
elsif (/^is_a:\s*GO:(\d+)/) { |
| 102 |
my $id2=int($1); |
| 103 |
print GOLINK join("\t",$c_id, $relcode{'is_a'}, $id2)."\n"; |
| 104 |
} |
| 105 |
elsif (/^subset:\s*(\S+)/) { |
| 106 |
my $subset=$1; |
| 107 |
print GOSET join("\t",$c_id, $subset)."\n"; |
| 108 |
} |
| 109 |
elsif (/^synonym:/) { |
| 110 |
my ($s,$t)=(m/\"(.+?)\"\s*(\w+)/); |
| 111 |
print STDERR "WARNING: parsing synonym & its type at $c_id, $_\n" unless $t; |
| 112 |
print GOSYN join("\t",$c_id, $s, $t)."\n" if $t; |
| 113 |
} |
| 114 |
elsif (m/^xref:\s*(\S+):(.+)$/) { |
| 115 |
my ($xdb, $xref)=($1,$2); |
| 116 |
$xref=~s/^\s+//;$xref=~s/\s+$//; |
| 117 |
print GOXREF join("\t",$c_id, $xdb, $xref)."\n"; |
| 118 |
} |
| 119 |
elsif (/^comment:\s*(.+)/) { |
| 120 |
my $comment=$1;$comment=~tr/\t\n/ /s; |
| 121 |
$c_comment=$comment; |
| 122 |
my @upds=($comment=~m/to\s*update\s*annotation.+?GO:(\d+)/g); |
| 123 |
foreach my $u_id (@upds) { |
| 124 |
print GOLINK join("\t",$c_id, 'U', int($u_id))."\n"; |
| 125 |
} |
| 126 |
} |
| 127 |
|
| 128 |
} #while |
| 129 |
goTerm() if $c_id; |
| 130 |
close(FGO); |
| 131 |
|
| 132 |
close(GOTERM); |
| 133 |
close(GOLINK); |
| 134 |
close(GOSYN); |
| 135 |
close(GOXREF); |
| 136 |
close(GOSET); |
| 137 |
|
| 138 |
|
| 139 |
#-- now run bcpin |
| 140 |
|
| 141 |
my $bcpcmd='bcpin -TI -b geanno@NEOSYBASE'; |
| 142 |
system($bcpcmd. ' go_term.bcp go_link.bcp go_synonyms.bcp go_xref.bcp go_subset.bcp') |
| 143 |
&& die("Error at bcpin!\n"); |
| 144 |
|
| 145 |
sub goTerm { |
| 146 |
print GOTERM join("\t",$c_id, $c_type, $c_name, $c_def, $c_obsolete, $c_comment)."\n"; |
| 147 |
} |
