Perl script that finds TreeFam transcripts that appear in the 'fam_genes' table of the TreeFam mysql database, but do not appear in the 'genes' table.

treefam_QC9.pl

#!/usr/local/bin/perl

# 
# Perl script treefam_QC9.pl
# Written by Avril Coghlan ([email protected])
# 3-Feb-09.
#
# This perl script finds TreeFam transcripts that appear in the fam_genes
# table, but do not appear in the 'genes' table.
#
# The command-line format is:
# % perl <treefam_QC9.pl> <release> 
#   where <release> is the release of the TreeFam database to use.
#
#
#------------------------------------------------------------------#

# CHECK IF THERE ARE THE CORRECT NUMBER OF COMMAND-LINE ARGUMENTS:

$num_args                  = $#ARGV + 1;
if ($num_args != 1)
{
   print "Usage of treefam_QC9.pl\n\n";
   print "perl treefam_QC9.pl <release>\n";
   print "where <release> is the release of the TreeFam database to use.\n";
   print "For example, >perl -w treefam_QC9.pl 7\n";
   exit;
}

#------------------------------------------------------------------#

# DECLARE MYSQL USERNAME AND HOST:
use DBI;

# FIND WHICH RELEASE OF THE TREEFAM DATABASE TO USE:

$release                   = $ARGV[0];

#------------------------------------------------------------------#

$database                  = "dbi:mysql:treefam_".$release.":db.treefam.org:3308";
$dbh                       = DBI->connect("$database", 'anonymous', '') || return;

# GET A LIST OF ALL THE TRANSCRIPTS IN THE 'genes' table:

$GENES                     = &read_genes_table($dbh);

# CHECK THAT ALL THE TRANSCRIPTS THAT WERE PUT IN FAMILIES APPEAR
# IN THE genes TABLE:

&read_transcripts_in_families($dbh,$GENES);

# NOW DISCONNECT FROM THE DATABASE:

$rc                        = $dbh->disconnect();
$rc                        = "";

#------------------------------------------------------------------#

print STDERR "FINISHED.\n";

#------------------------------------------------------------------#

# GET A LIST OF ALL THE TRANSCRIPTS IN THE 'genes' TABLE:

sub read_genes_table          
{
   my $dbh                 = $_[0];
   my $table_w;
   my $st;
   my $sth;
   my $rv;
   my @array;
   my $ID; 
   my %GENES               = ();

   $table_w                = 'genes';
   $st                     = "SELECT ID from $table_w"; 
   $sth                    = $dbh->prepare($st) or die "Cannot prepare $st: $dbh->errstr\n";
   $rv                     = $sth->execute or die "Cannot execute the query: $sth->errstr";
   if ($rv >= 1)
   {
      while ((@array) = $sth->fetchrow_array)    
      {
   $ID               = $array[0];
         $GENES{$ID}       = 1;
      }
   }

   print STDERR "Read in transcripts in the genes table\n";
   print "Read in transcripts in the genes table\n";

   return(\%GENES);
}

#------------------------------------------------------------------#

# CHECK THAT ALL TRANSCRIPTS PUT INTO FAMILIES APPEAR IN THE genes
# TABLE:

sub read_transcripts_in_families
{
   my $dbh                 = $_[0];
   my $GENES               = $_[1];
   my $table_w;
   my $st;
   my $sth;
   my $rv;
   my @array;
   my $ID;
   my $AC;
   my $FLAG;

   $table_w                = 'fam_genes';
   $st                     = "SELECT ID, AC, FLAG from $table_w"; 
   $sth                    = $dbh->prepare($st) or die "Cannot prepare $st: $dbh->errstr\n";
   $rv                     = $sth->execute or die "Cannot execute the query: $sth->errstr";
   if ($rv >= 1)
   {
      while ((@array) = $sth->fetchrow_array)    
      {
	 $ID               = $array[0];
         $AC               = $array[1];
         $FLAG             = $array[2];
         if (!($GENES->{$ID}))
         {
            print "WARNING: transcript $ID is in fam_genes table in $AC (FLAG $FLAG), but not in the genes table\n";
         }
      }
   }
}

#------------------------------------------------------------------#