# $Id: Geneid.pm 16123 2009-09-17 12:57:27Z cjfields $ # # Please direct questions and support issues to # # Cared for by Keith James # # Copyright Genome Research Ltd. # # You may distribute this module under the same terms as Perl itself # POD documentation - main docs before the code =head1 NAME Bio::Tools::Geneid - Results of one geneid run =head1 SYNOPSIS use Bio::Tools::Geneid; my $gid = Bio::Tools::Geneid(-file => "geneid.out"); while (my $gene = $gid->next_prediction) { my @transcripts = $gene->transcripts; foreach my $t (@transcripts) { my @exons = $t->exons; foreach my $e (@exons) { printf("Exon %d..%d\n", $e->start, $e->end); } } } =head1 DESCRIPTION This is the parser for the output of geneid by Enrique Blanco and Roderic Guigó (IMIM-UPF). See http://www1.imim.es/software/geneid. It relies on native geneid output format internally and will work with geneid versions 1.0 and 1.1. Currently this module supports only the default mode of operation which is to predict exons and assemble an optimal gene prediction. It takes either a file handle or a file name and returns a Bio::SeqFeature::Gene::GeneStructure object. =head1 FEEDBACK =head2 Mailing Lists User feedback is an integral part of the evolution of this and other Bioperl modules. Send your comments and suggestions preferably to one of the Bioperl mailing lists. Your participation is much appreciated. bioperl-l@bioperl.org - General discussion http://bioperl.org/wiki/Mailing_lists - About the mailing lists =head2 Support Please direct usage questions or support issues to the mailing list: I rather than to the module maintainer directly. Many experienced and reponsive experts will be able look at the problem and quickly address it. Please include a thorough description of the problem with code and data examples if at all possible. =head2 Reporting Bugs Report bugs to the Bioperl bug tracking system to help us keep track the bugs and their resolution. Bug reports can be submitted via the web: http://bugzilla.open-bio.org/ =head1 AUTHOR - Keith James Email: kdj@sanger.ac.uk =head1 APPENDIX The rest of the documentation details each of the object methods. Internal methods are usually preceded with a _ =cut package Bio::Tools::Geneid; use vars qw($SOURCE_TAG); use strict; use Bio::Tools::AnalysisResult; use Bio::SeqFeature::Generic; use Bio::SeqFeature::Gene::Exon; use Bio::SeqFeature::Gene::Transcript; use Bio::SeqFeature::Gene::GeneStructure; use base qw(Bio::Root::Root Bio::Root::IO); $SOURCE_TAG = 'geneid'; =head2 new Title : new Usage : $obj->new(-file = "new(-fh => \*GI); Function: Constructor for geneid wrapper. Takes either a file : or filehandle Returns : L =cut sub new { my($class, @args) = @_; my $self = $class->SUPER::new(@args); $self->_initialize_io(@args); return $self; } =head2 next_prediction Title : next_prediction Usage : while($gene = $geneid->next_prediction) { # do something } Function: Returns the gene structure prediction of the geneid result file. Call this method repeatedly until FALSE is returned. Returns : A Bio::SeqFeature::Gene::GeneStructure object Args : None =cut sub next_prediction { my ($self) = @_; my ($gene, $transcript, $current_gene_id); my $transcript_score = 0; my ($gene_id, $exon_type, $exon_start, $exon_end, $exon_score, $exon_strand, $start_phase, $end_phase, $start_sig_score, $end_sig_score, $coding_pot_score, $homol_score); while (defined($_ = $self->_readline)) { $self->debug($_); s/^\s+//; s/\s+$//; # We have a choice of geneid, gff or XML formats. The native # geneid format has more information than gff. However, we # then need to perform the hack of extracting the sequence ID # from the header of the embedded Fasta file which comes after # the exon data, as it is not stored elsewhere. Ack. # the new version of geneID includes the sequence ID in a slightly # different format and a new "or" statement was added below # also removed "unless defined $self->_target_id;" inorder to continue # generating new sequence IDs. if (/^>(\S+)\|GeneId/ or /^# Sequence (\S+)/) { my $target_id = $1; $self->_target_id($target_id); next; } next unless (/(Single|First|Internal|Terminal)/); my @fields = split(/\s+/, $_); # Grab gene_id from eol first as there are issues with # inconsistent whitespace in the AA coords field $gene_id = pop @fields; ($exon_type, $exon_start, $exon_end, $exon_score, $exon_strand, $start_phase, $end_phase, $start_sig_score, $end_sig_score, $coding_pot_score, $homol_score) = @fields[0..10]; if (! defined $current_gene_id) { # Starting the requested prediction $current_gene_id = $gene_id; $transcript_score = $exon_score; $gene = Bio::SeqFeature::Gene::GeneStructure->new(-source => $SOURCE_TAG); $transcript = Bio::SeqFeature::Gene::Transcript->new(-source => $SOURCE_TAG); $self->_add_exon($gene, $transcript, $exon_type, $exon_start, $exon_end, $exon_score, $exon_strand, $start_phase, $end_phase, $start_sig_score, $end_sig_score, $coding_pot_score, $homol_score); } elsif ($gene_id eq $current_gene_id) { # Still in requested prediction $transcript_score += $exon_score; $self->_add_exon($gene, $transcript, $exon_type, $exon_start, $exon_end, $exon_score, $exon_strand, $start_phase, $end_phase, $start_sig_score, $end_sig_score, $coding_pot_score, $homol_score); } else { # Found following prediction $self->_pushback($_); last; } } if (defined $gene) { $transcript->seq_id($self->_target_id); $transcript->score($transcript_score); $gene->add_transcript($transcript); $gene->seq_id($self->_target_id); foreach my $exon ($gene->exons) { $exon->seq_id($self->_target_id); } $self->_set_strand($gene); } return $gene; } =head2 _add_exon Title : _add_exon Usage : $obj->_add_exon($gene, $transcript, ... exon data ...) Function: Adds a new exon to both gene and transcript from the data : supplied as args Example : Returns : Nothing =cut sub _add_exon { my ($self, $gene, $transcript, $exon_type, $exon_start, $exon_end, $exon_score, $exon_strand, $start_phase, $end_phase, $start_sig_score, $end_sig_score, $coding_pot_score, $homol_score) = @_; $exon_type =~ s/First/Initial/; my $strand = $exon_strand eq '+' ? 1 : -1; my $exon = Bio::SeqFeature::Gene::Exon->new(-source => $SOURCE_TAG, -start => $exon_start, -end => $exon_end, -strand => $strand, -score => $exon_score); $exon->is_coding(1); $exon->add_tag_value("Type", $exon_type); $exon->add_tag_value('phase', $start_phase); $exon->add_tag_value('end_phase', $end_phase); $exon->add_tag_value('start_signal_score', $start_sig_score); $exon->add_tag_value('end_signal_score', $end_sig_score); $exon->add_tag_value('coding_potential_score', $coding_pot_score); $exon->add_tag_value('homology_score', $homol_score); $transcript->strand($strand) unless $transcript->strand != 0; $transcript->add_exon($exon, $exon_type); } =head2 _set_strand Title : _set_strand Usage : $obj->_set_strand($gene) Function: Sets the overall gene strand to the same strand as all : the exons if they are all on the same strand, or to strand 0 : if the exons are on different strands. Example : Returns : Nothing =cut sub _set_strand { my ($self, $gene) = @_; my $fwd = 0; my $rev = 0; my @exons = $gene->exons; foreach my $exon (@exons) { my $strand = $exon->strand; if ($strand == 1) { $fwd++; } elsif ($strand == -1) { $rev++; } } if ($#exons == $fwd) { $gene->strand(1); } elsif ($#exons == $rev) { $gene->strand(-1); } else { $gene->strand(0); } return $gene; } =head2 _target_id Title : _target_id Usage : $obj->_target_id Function: get/set for genomic sequence id Example : Returns : A target ID =cut sub _target_id { my ($self,$val) = @_; if ($val) { $self->{'_target_id'} = $val; } return $self->{'_target_id'}; } 1;