From a4b9214b7939fbe990f2373684dfd7aa1f9e77e2 Mon Sep 17 00:00:00 2001
From: Don Armstrong <don@donarmstrong.com>
Date: Sat, 27 Oct 2007 00:14:42 +0000
Subject: [PATCH]  * Fix genecard and harvester parsers  * Add headers for
 results_to_table  * Call results_to_table in function2gene

git-svn-id: file:///srv/svn/function2gene/trunk@15 a0738b58-4706-0410-8799-fb830574a030
---
 bin/function2gene           | 13 +++++++++++--
 bin/parse_genecard_results  | 37 ++++++++++++++++++++++++-------------
 bin/parse_harvester_results |  7 +++++++
 bin/results_to_table        |  6 ++++--
 4 files changed, 46 insertions(+), 17 deletions(-)

diff --git a/bin/function2gene b/bin/function2gene
index 07dd2ae..ed9a38c 100755
--- a/bin/function2gene
+++ b/bin/function2gene
@@ -56,7 +56,8 @@ databases.]
 =item B<--restart-at>
 
 If you need to restart the process at a particular state (which has
-already been completed) specify this option.
+already been completed) specify this option. Valid values are get,
+parse, or combine.
 
 =item B<--debug, -d>
 
@@ -78,6 +79,10 @@ Display this manual.
    echo 'transferrin' > keywords.txt
    function2gene --keywords keywords.txt --results keyword_results
 
+   # reparse the results
+   function2gene --keywords keywords.txt --results keyword_results \
+       --restart-at parse
+
 =cut
 
 
@@ -287,7 +292,11 @@ if ($actions{combine}) {
 	  $state{done_keywords}{combined}{$db}{$keyword} = 1;
      }
      save_state(\%state);
-     ADVISE("Finished; results in $options{results}/combined_results");
+     write_command_to_file('combined_results_table.txt',
+			   "$base_dir/results_to_table",
+			   'combined_results.txt',
+			  );
+     ADVISE("Finished; results in $options{results}/combined_results.txt");
 }
 else {
      ADVISE('Nothing to do. [Perhaps you wanted --restart-at?]');
diff --git a/bin/parse_genecard_results b/bin/parse_genecard_results
index 4b7d514..b423719 100755
--- 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
      $results[DBNAME] = 'genecard';
diff --git a/bin/parse_harvester_results b/bin/parse_harvester_results
index 30fc011..e520661 100755
--- a/bin/parse_harvester_results
+++ b/bin/parse_harvester_results
@@ -142,6 +142,13 @@ while ($_ = $dir->read) {
 	  ($results[NAME]) = $result =~ m&<TR>\s*<TD\s*BGCOLOR="\#FEFE99"\s*VALIGN="top"\s*NOWRAP>Entry\s*name</TD>\s*
 					  <TD\s*VALIGN="top"\s*COLSPAN="5">\s*<b>\s*([^<]+?)\s*</b></TD>\s*</TR>&xis;
      }
+     if (not defined $results[NAME]) {
+	  ($results[NAME]) = $result =~ m{<TITLE>[^:]+:\s*[^\*]+\*[^\*]+\*\s*([^-]+)}xis;
+	  $results[NAME] =~ s/\s*$// if defined $results[NAME];
+	  $results[NAME] =~ s/^\s*$// if defined $results[NAME];
+	  $results[NAME] =~ s/\d+\s*kDa\s*protein// if defined $results[NAME];
+	  $results[NAME] =~ s/\s*similar to .+// if defined $results[NAME];
+     }
 
      $results[NAME] ||= 'NO NAME';
      $results[NAME] =~ s/_HUMAN//;
diff --git a/bin/results_to_table b/bin/results_to_table
index 04f2f6d..24c9400 100755
--- a/bin/results_to_table
+++ b/bin/results_to_table
@@ -75,12 +75,14 @@ while (<$search_results_fh>) {
      $databases{total}{total}{count}++;
 }
 
-our ($keyword,$gct,$hvt,$nct,$t);
+our ($keyword,$gct,$hvt,$nct,$t) = ('Keyword','GeneCards','Harvester','NCBI','Total');
      format STDOUT =
 @<<<<<<<<<<<<<<<<<<<<<< & @>>>>>>>>>> & @>>>>>>>>>> & @>>>>>>>>>> & @>>>>>>>>>> \\
 $keyword,                 $gct,     $hvt,     $nct,     $t
 .
-for$keyword (sort keys %terms) {
+write;
+
+for $keyword (sort keys %terms) {
      ($gct,$hvt,$nct,$t) =
 	  map {
 	       if (not defined $_) {
-- 
2.39.2