Ok then you can do it this way:

my %transcripts;
{
   open(my $transcripts_fh, "<", $transcripts_qfn)
      or die("Can't open \"$transcripts_qfn\": $!\n");
   while (<$transcripts_fh>) {
      chomp;
      my @refs = split(/\t/, $_);
      my ($ref_chr, $ref_strand) = @refs[0, 6];
      my $values= {start => $refs[3], end => $refs[4], info => $refs[8
+]};
      push @{ $transcripts{$ref_chr}{$ref_strand} }, $values;
   }    
}
# You should really debug the output at least once with Data::Dumper t
+o see how it looks like
[download]

Now you use that like this:

my $transcripts_array = $transcripts{$chr}{$strand}; # might need a be
+tter name
for my $transcript (@$transcripts_array)
{
  my $start = $transcript->{start};
  ...
}
[download]

This is great! Thanks a ton! As a test I tried my old script vs the better one with your method on just 10 lines on my not high-end work computer:

Elapsed time with your script: 00:00:00.959781

Elapsed time with my original: 00:00:02.324184

Multiply this difference by a few hundred thousand for the complete input files, and you can really note the improvement.

If you are still implementing threads: Don't forget to use flock while writing to file inside the threads, something like this:

use threads;
use Fcntl qw(:flock SEEK_END);


      if ($end >= $ref->{start} && $end <= $ref->{end}) {
    lock($out_fh);
    say $out_fh join("\t", $ID, $strand, $chr, $start, $end, $sequence
+, $numPositions, $mismatches||"", $ref->{info});
    unlock($out_fh);
      }


sub lock {
  my ($fh) = @_;
  flock($fh, LOCK_EX) or die $!;
  seek($fh, 0, SEEK_END) or die $!;
}
sub unlock {
  my ($fh) = @_;
  flock($fh, LOCK_UN) or die $!;
}
[download]

Alternatively, use shared objects (collect it in an array or hash) and write it down later.

Great Eily, thanks. I will test this. But in the meantime just to answer the question, it is possible that an entry in the input file could map to multiple transcripts.