* Fix genecard and harvester parsers

[function2gene.git] / bin / parse_genecard_results

diff --git a/bin/parse_genecard_results b/bin/parse_genecard_results
index 4b7d514d7df0b5352a4fb71aa50f0d8cfd55365e..b42371957a229b4423cd8b89ac152eb2094f7134 100755 (executable)
--- a/bin/parse_genecard_results
+++ b/bin/parse_genecard_results
@@ -70,6 +70,9 @@ BEGIN{
 use IO::File;
 use IO::Dir;
 
+use HTML::TreeBuilder;
+use HTML::ElementTable;
+
 my %options = (debug    => 0,
               help     => 0,
               man      => 0,
@@ -139,29 +142,26 @@ while ($_ = $dir->read) {
      $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/\n//g}
-                      map {s#\s*\s*</a>(?:</td><td>\s*)?\s*# #g; $_;}
-                      $result =~ m{(GO:\d+\s*</a>(?:</td><td>\s*)?.+?)(?:</td><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);
      $results[FUNCTION] ||= 'NO FUNCTION';
@@ -171,8 +171,19 @@ while ($_ = $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';