* Fix genecard and harvester parsers

[function2gene.git] / bin / parse_genecard_results

diff --git a/bin/parse_genecard_results b/bin/parse_genecard_results
index 4c00d9b47b85b44ef07adb66042f224cd6513e45..b42371957a229b4423cd8b89ac152eb2094f7134 100755 (executable)
--- a/bin/parse_genecard_results
+++ b/bin/parse_genecard_results
@@ -70,16 +70,20 @@ BEGIN{
 use IO::File;
 use IO::Dir;
 
-# XXX parse config file
+use HTML::TreeBuilder;
+use HTML::ElementTable;
 
 my %options = (debug    => 0,
               help     => 0,
               man      => 0,
               dir      => '.',
               keyword  => undef,
+              keywords => 0,
              );
 
-GetOptions(\%options,'keyword|k=s','dir|D=s','debug|d+','help|h|?','man|m');
+GetOptions(\%options,'keyword|k=s','dir|D=s','debug|d+','help|h|?','man|m',
+          'keywords',
+         );
 
 
 pod2usage() if $options{help};
@@ -99,6 +103,13 @@ use constant {NAME        => 0,
              FILENAME    => 8,
             };
 
+if ($options{keywords}) {
+     if (@ARGV != 1) {
+         pod2usage("If the --keywords option is used, exactly one argument (the keyword) must be passed");
+     }
+     $options{dir} = "$ARGV[0]_results_genecard";
+}
+
 if (not -d $options{dir}) {
      die "$options{dir} does not exist or is not a directory";
 }
@@ -120,40 +131,38 @@ while ($_ = $dir->read) {
      my @results;
 
      # Find gene name
-     ($results[NAME]) = $result =~ m&(?:Lean|Gene)Card\s+for\s+(?:(?:disorder\s+locus|uncategorized|
-                                    hugo\s*reserved\s*symbol|cluster|
-                                    potentially\s*expressed\s*sequence)|(?:predicted\s+|pseudo|rna\s+|)gene)
-                                    \s*(?:with\s*support\s*|)<FONT\s+COLOR=\"[^\"]+\">\s*<FONT\s+SIZE=\+2>\s*([^\s]+)\s*&xis;
+     ($results[NAME]) = $result =~ m{(?:Lean|Gene)Card\s+for\s+[^<]+<FONT[^>]+>\s*([^<]+)}xis;
 
      $results[NAME] ||= 'NO NAME';
      # Find REF SEQ number
-     ($results[REFSEQ]) = $result =~ m|http://www.ncbi.nlm.nih.gov/entrez/query.fcgi\?
-                              cmd=Search\&db=nucleotide\&doptcmdl=GenBank\&term=([^\"]+)\"|xis;
+     ($results[REFSEQ]) = $result =~ m{http://www.ncbi.nlm.nih.gov/entrez/query.fcgi\?
+                                      (?:cmd=Search\&db=nucleotide|db=nucleotide\&cmd=search)
+                                      \&doptcmdl=GenBank\&term=([^\"]+)\"}xis;
 
      $results[REFSEQ] ||= 'NO REFSEQ';
 
      # Find Gene Location
-     ($results[LOCATION]) = $result =~ m&<I>LocusLink\s+cytogenetic\s+band:</I><b>\s+
-                                        <a\s+href="[^\"]+"\s+target\s+=\s+"aaa">\s*([^\<]+?)\s*</a>&xis;
+     ($results[LOCATION]) = $result =~ m{cytogenetic\s+band:</I><b>\s+
+                                        <a\s+href="[^\"]+"\s+target\s*=\s*"aaa"[^>]*>\s*([^\<]+?)\s*</a>}xis;
 
      $results[LOCATION] ||= 'NO LOCATION';
 
      # Find gene aliases
-     my ($alias_table) = $result =~ m|<b>Aliases and Descriptions</b>(.+?)</TR>|is;
-     $alias_table ||='';
-
-     my @gene_aliases = $alias_table =~ m|<li>\s*([^\(]{0,20}?)\s*\(<FONT|gis;
+     my ($alias_table) = $result =~ m{(<table[^>]+><tr><th[^>]+>Aliases.+?</table>)}is;
+     $alias_table ||= '';
+     my @gene_aliases = map {s/\s*$//; $_;} $alias_table =~ m{<td(?: nowrap)?>\s*([^<]+)<}gis;
 
      $results[ALIAS] = join('; ', @gene_aliases);
      $results[ALIAS] ||= 'NO ALIASES';
 
      # Find gene function(s)
 
-     # Swiss prot functions
-     my @functions = $result =~ m&<li><b>Function:</b>\s+(.+?)(?:<li>)|(?:</ul>)&gis;
-
+     my @functions;
      # GO Functions
-     push @functions, (map {s#\s*</a>\s*# #g; $_;} $result =~ m&(GO:\d+\s*</a>.+?)(?:<dd>|<p>)&gis);
+     push @functions, (map {s/\n//g; $_;}
+                      map {s#\s*</a>(?:</td><td>\s*)?\s*# #g; $_;}
+                      $result =~ m{(GO:\d+\s*</a>(?:</td><td>\s*)?.+?)(?:</font>|</td>|<dd>|<p>)}gis
+                     );
      $results[FUNCTION] = join('; ', map {(defined $_)?($_):()} @functions);
      $results[FUNCTION] ||= 'NO FUNCTION';
 
@@ -162,8 +171,19 @@ while ($_ = $dir->read) {
 
      $results[KEYWORD] ||= 'NO KEYWORD';
 
+     # Swiss prot functions
+     my @description = (map {s/<[^>]+>/ /g;
+                          s/\s+/ /g;
+                          $_;
+                     }
+                     $result =~ m{<(?:dd|li)><b>Function(?::</b>|</b>:)\s+
+                                  (.+?)<(?:/dd|li)>}xgis
+                    );
+
      # Figure out what the description is
-     $results[DESCRIPTION] = '';
+     $results[DESCRIPTION] = join('; ',
+                                 map {(defined $_)?($_):()}
+                                 @description);
 
      # Database searched
      $results[DBNAME] = 'genecard';