From ce19e68064d36484e4932fbebd10810c2a6502bd Mon Sep 17 00:00:00 2001
From: Patrick Meidl <pm2@sanger.ac.uk>
Date: Mon, 28 Apr 2008 20:59:53 +0000
Subject: [PATCH] plugin architecture for InternalIdMapper
---
misc-scripts/id_mapping/default.conf | 23 +-
misc-scripts/id_mapping/id_mapping.pl | 3 +
misc-scripts/id_mapping/run.pl | 3 +
.../Bio/EnsEMBL/IdMapping/InternalIdMapper.pm | 621 ++++--------------
.../IdMapping/InternalIdMapper/BaseMapper.pm | 250 +++++++
.../InternalIdMapper/EnsemblExonGeneric.pm | 91 +++
.../InternalIdMapper/EnsemblGeneGeneric.pm | 187 ++++++
.../EnsemblTranscriptGeneric.pm | 276 ++++++++
8 files changed, 971 insertions(+), 483 deletions(-)
create mode 100644 modules/Bio/EnsEMBL/IdMapping/InternalIdMapper/BaseMapper.pm
create mode 100644 modules/Bio/EnsEMBL/IdMapping/InternalIdMapper/EnsemblExonGeneric.pm
create mode 100644 modules/Bio/EnsEMBL/IdMapping/InternalIdMapper/EnsemblGeneGeneric.pm
create mode 100644 modules/Bio/EnsEMBL/IdMapping/InternalIdMapper/EnsemblTranscriptGeneric.pm
diff --git a/misc-scripts/id_mapping/default.conf b/misc-scripts/id_mapping/default.conf
index 2d90cd045c..8d5b95c597 100644
--- a/misc-scripts/id_mapping/default.conf
+++ b/misc-scripts/id_mapping/default.conf
@@ -5,7 +5,7 @@ dry_run = 0
loglevel = DEBUG
; paths
-basedir = /lustre/work1/ensembl/pm2/idmapping/perl/2008-04-22c
+basedir = /lustre/work1/ensembl/pm2/idmapping/perl/2008-04-28
; prepend this path to your 'log' parameter
; will default to "$basedir/log" if not set
@@ -27,7 +27,7 @@ targetdbname = pm2_pan_troglodytes_core_41_21
; caching
;cache_method = build_cache_all
build_cache_auto_threshold = 100
-build_cache_concurrent_jobs = 200
+build_cache_concurrent_jobs = 200
; limit
;region = chromosome:CHIMP1A:1:1:2000000:1
@@ -50,6 +50,25 @@ transcript_score_threshold = 0
synteny_rescore_jobs = 20
;lsf_opt_synteny_rescore =
+; InternalIdMapper
+;plugin_internal_id_mappers_gene = \
+; Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblGeneGeneric::init_basic,\
+; Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblGeneGeneric::synteny,\
+; Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblGeneGeneric::best_transcript,\
+; Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblGeneGeneric::biotype,\
+; Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblGeneGeneric::internal_id
+
+;plugin_internal_id_mappers_transcript = \
+; Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblTranscriptGeneric::init_basic,\
+; Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblTranscriptGeneric::non_exact_translation,\
+; Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblTranscriptGeneric::mapped_gene,\
+; Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblTranscriptGeneric::internal_id,\
+; Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblTranscriptGeneric::single_gene
+
+;plugin_internal_id_mappers_exon = \
+; Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblExonGeneric::init_basic,\
+; Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblExonGeneric::mapped_transcript
+
; StableIdMapper
mapping_types = gene,transcript,translation,exon
;plugin_stable_id_generator = Bio::EnsEMBL::IdMapping::StableIdGenerator::EnsemblGeneric
diff --git a/misc-scripts/id_mapping/id_mapping.pl b/misc-scripts/id_mapping/id_mapping.pl
index 6030851bf0..f1b4903d2b 100755
--- a/misc-scripts/id_mapping/id_mapping.pl
+++ b/misc-scripts/id_mapping/id_mapping.pl
@@ -87,6 +87,9 @@ $conf->parse_options(
'exonerate_jobs|exoneratejobs=i' => 0,
'exonerate_bytes_per_job|exoneratebytesperjob=f' => 0,
'exonerate_extra_params|exonerateextraparams=s' => 0,
+ 'plugin_internal_id_mappers_gene=s@' => 0,
+ 'plugin_internal_id_mappers_transcript=s@' => 0,
+ 'plugin_internal_id_mappers_exon=s@' => 0,
'mapping_types=s@' => 1,
'plugin_stable_id_generator=s' => 0,
'upload_events|uploadevents=s' => 0,
diff --git a/misc-scripts/id_mapping/run.pl b/misc-scripts/id_mapping/run.pl
index ad367d3ad0..f764352b89 100755
--- a/misc-scripts/id_mapping/run.pl
+++ b/misc-scripts/id_mapping/run.pl
@@ -96,6 +96,9 @@ $conf->parse_options(
'exonerate_jobs|exoneratejobs=i' => 0,
'exonerate_bytes_per_job|exoneratebytesperjob=f' => 0,
'exonerate_extra_params|exonerateextraparams=s' => 0,
+ 'plugin_internal_id_mappers_gene=s@' => 0,
+ 'plugin_internal_id_mappers_transcript=s@' => 0,
+ 'plugin_internal_id_mappers_exon=s@' => 0,
'mapping_types=s@' => 1,
'plugin_stable_id_generator=s' => 0,
'upload_events|uploadevents=s' => 0,
diff --git a/modules/Bio/EnsEMBL/IdMapping/InternalIdMapper.pm b/modules/Bio/EnsEMBL/IdMapping/InternalIdMapper.pm
index bd2fe1b497..449b584369 100644
--- a/modules/Bio/EnsEMBL/IdMapping/InternalIdMapper.pm
+++ b/modules/Bio/EnsEMBL/IdMapping/InternalIdMapper.pm
@@ -37,7 +37,7 @@ use Bio::EnsEMBL::IdMapping::BaseObject;
our @ISA = qw(Bio::EnsEMBL::IdMapping::BaseObject);
use Bio::EnsEMBL::Utils::Exception qw(throw warning);
-use Bio::EnsEMBL::Utils::ScriptUtils qw(path_append);
+use Bio::EnsEMBL::Utils::ScriptUtils qw(inject path_append);
use Bio::EnsEMBL::IdMapping::Entry;
use Bio::EnsEMBL::IdMapping::MappingList;
use Bio::EnsEMBL::IdMapping::SyntenyFramework;
@@ -93,110 +93,44 @@ sub map_genes {
# create gene mappings
$self->logger->info("No gene mappings found. Will calculate them now.\n");
- #
- # basic mapping
- #
- $self->logger->info("Basic gene mapping...\n", 0, 'stamped');
-
- my $mappings0 = $self->basic_mapping($gene_scores, 'gene_mappings0');
-
- my $gene_scores1 = $gsb->create_shrinked_matrix($gene_scores, $mappings0,
- 'gene_matrix1');
-
-
- #
- # build the synteny from unambiguous mappings
- #
- unless ($gene_scores1->loaded) {
- $self->logger->info("Synteny Framework building...\n", 0, 'stamped');
- my $sf = Bio::EnsEMBL::IdMapping::SyntenyFramework->new(
- -DUMP_PATH => $dump_path,
- -CACHE_FILE => 'synteny_framework.ser',
- -LOGGER => $self->logger,
- -CONF => $self->conf,
- -CACHE => $self->cache,
- );
- $sf->build_synteny($mappings0);
-
- # use it to rescore the genes
- $self->logger->info("\nSynteny assisted mapping...\n", 0, 'stamped');
- $gene_scores1 = $sf->rescore_gene_matrix_lsf($gene_scores1);
-
- # checkpoint
- $gene_scores1->write_to_file;
- }
-
- my $mappings1 = $self->basic_mapping($gene_scores1, 'gene_mappings1');
-
- my $gene_scores2 = $gsb->create_shrinked_matrix($gene_scores1, $mappings1,
- 'gene_matrix2');
-
+ # determine which plugin methods to run
+ my @default_plugins = (qw(
+ Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblGeneGeneric::init_basic
+ Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblGeneGeneric::synteny
+ Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblGeneGeneric::best_transcript
+ Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblGeneGeneric::biotype
+ Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblGeneGeneric::internal_id
+ ));
+
+ my @plugins = $self->conf->param('plugin_internal_id_mappers_gene');
+ @plugins = @default_plugins unless (defined($plugins[0]));
+
+ my $new_mappings = Bio::EnsEMBL::IdMapping::MappingList->new(
+ -DUMP_PATH => $dump_path,
+ -CACHE_FILE => 'gene_mappings0.ser',
+ );
+ my @mappings = ();
+ my $i = 0;
#
- # rescore with simple scoring function and try again
+ # run the scoring chain
#
- $self->logger->info("Retry with simple best transcript score...\n", 0, 'stamped');
-
- unless ($gene_scores2->loaded) {
- $gsb->simple_gene_rescore($gene_scores2, $transcript_scores);
- $gene_scores2->write_to_file;
- }
-
- my $mappings2 = $self->basic_mapping($gene_scores2, 'gene_mappings2');
-
- my $gene_scores3 = $gsb->create_shrinked_matrix($gene_scores2, $mappings2,
- 'gene_matrix3');
+ foreach my $plugin (@plugins) {
+ ($gene_scores, $new_mappings) = $self->delegate_to_plugin($plugin, $i++,
+ $gsb, $new_mappings, $gene_scores, $transcript_scores);
-
- #
- # rescore by penalising scores between genes with different biotypes
- #
- $self->logger->info("Retry with biotype disambiguation...\n", 0, 'stamped');
-
- unless ($gene_scores3->loaded) {
- $gsb->biotype_gene_rescore($gene_scores3);
- $gene_scores3->write_to_file;
+ push(@mappings, $new_mappings);
}
- my $mappings3 = $self->basic_mapping($gene_scores3, 'gene_mappings3');
-
- my $gene_scores4 = $gsb->create_shrinked_matrix($gene_scores3, $mappings3,
- 'gene_matrix4');
-
-
- #
- # selectively rescore by penalising scores between genes with different
- # internalIDs
- #
- $self->logger->info("Retry with internalID disambiguation...\n", 0, 'stamped');
-
- unless ($gene_scores4->loaded) {
- $gsb->internal_id_rescore($gene_scores4);
- $gene_scores4->write_to_file;
- }
-
- my $mappings4 = $self->basic_mapping($gene_scores4, 'gene_mappings4');
-
- my $remaining_gene_scores = $gsb->create_shrinked_matrix(
- $gene_scores4, $mappings4, 'remaining_gene_matrix');
-
-
- #
# report remaining ambiguities
- #
- $self->logger->info($remaining_gene_scores->get_source_count.
+ $self->logger->info($gene_scores->get_source_count.
" source genes are ambiguous with ".
- $remaining_gene_scores->get_target_count." target genes.\n\n");
-
- $self->log_ambiguous($remaining_gene_scores, 'gene');
+ $gene_scores->get_target_count." target genes.\n\n");
+ $self->log_ambiguous($gene_scores, 'gene');
- #
# merge mappings and write to file
- #
- $mappings->add_all($mappings0, $mappings1, $mappings2, $mappings3,
- $mappings4);
-
+ $mappings->add_all(@mappings);
$mappings->write_to_file;
if ($self->logger->loglevel eq 'debug') {
@@ -258,103 +192,44 @@ sub map_transcripts {
# create transcript mappings
$self->logger->info("No transcript mappings found. Will calculate them now.\n");
- #
- # basic mapping
- #
- $self->logger->info("Basic transcript mapping...\n", 0, 'stamped');
-
- my $mappings0 = $self->basic_mapping($transcript_scores,
- 'transcript_mappings0');
-
- my $transcript_scores1 = $tsb->create_shrinked_matrix(
- $transcript_scores, $mappings0, 'transcript_matrix1');
-
-
- #
- # handle cases with exact match but different translation
- #
- $self->logger->info("Exact Transcript non-exact Translation...\n", 0, 'stamped');
-
- unless ($transcript_scores1->loaded) {
- $tsb->different_translation_rescore($transcript_scores1);
- $transcript_scores1->write_to_file;
- }
-
- my $mappings1 = $self->basic_mapping($transcript_scores1,
- 'transcript_mappings1');
-
- my $transcript_scores2 = $tsb->create_shrinked_matrix(
- $transcript_scores1, $mappings1, 'transcript_matrix2');
-
+ # determine which plugin methods to run
+ my @default_plugins = (qw(
+ Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblTranscriptGeneric::init_basic
+ Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblTranscriptGeneric::non_exact_translation
+ Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblTranscriptGeneric::mapped_gene
+ Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblTranscriptGeneric::internal_id
+ Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblTranscriptGeneric::single_gene
+ ));
+
+ my @plugins = $self->conf->param('plugin_internal_id_mappers_transcript');
+ @plugins = @default_plugins unless (defined($plugins[0]));
+
+ my $new_mappings = Bio::EnsEMBL::IdMapping::MappingList->new(
+ -DUMP_PATH => $dump_path,
+ -CACHE_FILE => 'transcript_mappings0.ser',
+ );
+ my @mappings = ();
+ my $i = 0;
#
- # reduce score for mappings of transcripts which do not belong to mapped
- # genes
+ # run the scoring chain
#
- $self->logger->info("Transcripts in mapped genes...\n", 0, 'stamped');
-
- unless ($transcript_scores2->loaded) {
- $tsb->non_mapped_gene_rescore($transcript_scores2, $gene_mappings);
- $transcript_scores2->write_to_file;
- }
-
- my $mappings2 = $self->basic_mapping($transcript_scores2,
- 'transcript_mappings2');
-
- my $transcript_scores3 = $tsb->create_shrinked_matrix(
- $transcript_scores2, $mappings2, 'transcript_matrix3');
-
+ foreach my $plugin (@plugins) {
+ ($transcript_scores, $new_mappings) = $self->delegate_to_plugin($plugin,
+ $i++, $tsb, $new_mappings, $transcript_scores, $gene_mappings);
- #
- # selectively rescore by penalising scores between transcripts with
- # different internalIDs
- #
- $self->logger->info("Retry with internalID disambiguation...\n", 0, 'stamped');
-
- unless ($transcript_scores3->loaded) {
- $tsb->internal_id_rescore($transcript_scores3);
- $transcript_scores3->write_to_file;
+ push(@mappings, $new_mappings);
}
- my $mappings3 = $self->basic_mapping($transcript_scores3,
- 'transcript_mappings3');
-
- my $transcript_scores4 = $tsb->create_shrinked_matrix(
- $transcript_scores3, $mappings3, 'transcript_matrix4');
-
-
- #
- # handle ambiguities between transcripts in single genes
- #
- $self->logger->info("Transcripts in single genes...\n", 0, 'stamped');
-
- unless ($transcript_scores4->loaded) {
- $transcript_scores4->write_to_file;
- }
-
- my $mappings4 = $self->same_gene_transcript_mapping($transcript_scores4,
- 'transcript_mappings4');
-
- my $remaining_transcript_scores = $tsb->create_shrinked_matrix(
- $transcript_scores4, $mappings4, 'transcript_matrix5');
-
-
- #
# report remaining ambiguities
- #
- $self->logger->info($remaining_transcript_scores->get_source_count.
+ $self->logger->info($transcript_scores->get_source_count.
" source transcripts are ambiguous with ".
- $remaining_transcript_scores->get_target_count." target transcripts.\n\n");
+ $transcript_scores->get_target_count." target transcripts.\n\n");
- $self->log_ambiguous($remaining_transcript_scores, 'transcript');
+ $self->log_ambiguous($transcript_scores, 'transcript');
-
- #
# merge mappings and write to file
- #
- $mappings->add_all($mappings0, $mappings1, $mappings2, $mappings3,
- $mappings4);
-
+ $mappings->add_all(@mappings);
$mappings->write_to_file;
if ($self->logger->loglevel eq 'debug') {
@@ -417,49 +292,41 @@ sub map_exons {
# create exon mappings
$self->logger->info("No exon mappings found. Will calculate them now.\n");
- #
- # basic mapping
- #
- $self->logger->info("Basic exon mapping...\n", 0, 'stamped');
+ # determine which plugin methods to run
+ my @default_plugins = (qw(
+ Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblExonGeneric::init_basic
+ Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblExonGeneric::mapped_transcript
+ ));
- my $mappings0 = $self->basic_mapping($exon_scores, 'exon_mappings0');
+ my @plugins = $self->conf->param('plugin_internal_id_mappers_exon');
+ @plugins = @default_plugins unless (defined($plugins[0]));
- my $exon_scores1 = $esb->create_shrinked_matrix( $exon_scores, $mappings0,
- 'exon_matrix1');
-
+ my $new_mappings = Bio::EnsEMBL::IdMapping::MappingList->new(
+ -DUMP_PATH => $dump_path,
+ -CACHE_FILE => 'exon_mappings0.ser',
+ );
+ my @mappings = ();
+ my $i = 0;
#
- # reduce score for mappings of exons which do not belong to mapped
- # transcripts
+ # run the scoring chain
#
- $self->logger->info("Exons in mapped transcripts...\n", 0, 'stamped');
-
- unless ($exon_scores1->loaded) {
- $esb->non_mapped_transcript_rescore($exon_scores1, $transcript_mappings);
- $exon_scores1->write_to_file;
- }
-
- my $mappings1 = $self->basic_mapping($exon_scores1, 'exon_mappings1');
-
- my $remaining_exon_scores = $esb->create_shrinked_matrix(
- $exon_scores1, $mappings1, 'exon_matrix2');
+ foreach my $plugin (@plugins) {
+ ($exon_scores, $new_mappings) = $self->delegate_to_plugin($plugin, $i++,
+ $esb, $new_mappings, $exon_scores);
+ push(@mappings, $new_mappings);
+ }
- #
# report remaining ambiguities
- #
- $self->logger->info($remaining_exon_scores->get_source_count.
+ $self->logger->info($exon_scores->get_source_count.
" source exons are ambiguous with ".
- $remaining_exon_scores->get_target_count." target exons.\n\n");
+ $exon_scores->get_target_count." target exons.\n\n");
- $self->log_ambiguous($remaining_exon_scores, 'exon');
+ $self->log_ambiguous($exon_scores, 'exon');
-
- #
# merge mappings and write to file
- #
- $mappings->add_all($mappings0, $mappings1);
-
+ $mappings->add_all(@mappings);
$mappings->write_to_file;
if ($self->logger->loglevel eq 'debug') {
@@ -475,6 +342,10 @@ sub map_exons {
}
+#
+# this is not implemented as a plugin, since a) it's too simple and b) it's
+# tied to transcripts so there are no translation scores or score builder.
+#
sub map_translations {
my $self = shift;
my $transcript_mappings = shift;
@@ -557,304 +428,92 @@ sub map_translations {
}
-#
-# find the highest unambiguous score for all sources and targets in a scoring
-# matrix
-#
-sub basic_mapping {
+sub delegate_to_plugin {
my $self = shift;
- my $matrix = shift;
- my $mapping_name = shift;
+ my $plugin = shift;
+ my $num = shift;
+ my $score_builder = shift;
+ my $mappings = shift;
+ my $scores = shift;
# argument checks
- unless ($matrix and
- $matrix->isa('Bio::EnsEMBL::IdMapping::ScoredMappingMatrix')) {
- throw('Need a Bio::EnsEMBL::IdMapping::ScoredMappingMatrix.');
+ unless ($score_builder and
+ $score_builder->isa('Bio::EnsEMBL::IdMapping::ScoreBuilder')) {
+ throw('Need a Bio::EnsEMBL::IdMapping::ScoreBuilder.');
}
- throw('Need a name for serialising the mapping.') unless ($mapping_name);
-
- # Create a new MappingList object. Specify AUTO_LOAD to load serialised
- # existing mappings if found
- my $dump_path = path_append($self->conf->param('basedir'), 'mapping');
-
- my $mappings = Bio::EnsEMBL::IdMapping::MappingList->new(
- -DUMP_PATH => $dump_path,
- -CACHE_FILE => "${mapping_name}.ser",
- -AUTO_LOAD => 1,
- );
-
- # checkpoint test: return a previously stored MappingList
- if ($mappings->loaded) {
- $self->logger->info("Read existing mappings from ${mapping_name}.ser.\n");
- return $mappings;
+ unless ($mappings and
+ $mappings->isa('Bio::EnsEMBL::IdMapping::MappingList')) {
+ throw('Need a Bio::EnsEMBL::IdMapping::MappingList.');
}
-
- my $sources_done = {};
- my $targets_done = {};
-
- # sort scoring matrix entries by descending score
- my @sorted_entries = sort { $b->score <=> $a->score }
- @{ $matrix->get_all_Entries };
-
- # debug
- my $idx = substr($mapping_name, -1);
-
- while (my $entry = shift(@sorted_entries)) {
-
- #$self->logger->debug("\nxxx$idx ".$entry->to_string." ");
-
- # we already found a mapping for either source or target
- next if ($sources_done->{$entry->source} or
- $targets_done->{$entry->target});
-
- #$self->logger->debug('d');
-
- # there's a better mapping for either source or target
- next if ($self->higher_score_exists($entry, $matrix, $sources_done,
- $targets_done));
-
- #$self->logger->debug('h');
-
- # check for ambiguous mappings; they are dealt with later
- my $other_sources = [];
- my $other_targets = [];
-
- if ($self->ambiguous_mapping($entry, $matrix, $other_sources, $other_targets)) {
- #$self->logger->debug('a');
-
- $other_sources = $self->filter_sources($other_sources, $sources_done);
- $other_targets = $self->filter_targets($other_targets, $targets_done);
-
- next if (scalar(@$other_sources) or scalar(@$other_targets));
- }
-
- #$self->logger->debug('A');
-
- # this is the best mapping, add it
- $mappings->add_Entry($entry);
-
- $sources_done->{$entry->source} = 1;
- $targets_done->{$entry->target} = 1;
+
+ unless ($scores and
+ $scores->isa('Bio::EnsEMBL::IdMapping::ScoredMappingMatrix')) {
+ throw('Need a Bio::EnsEMBL::IdMapping::ScoredMappingMatrix.');
}
- # create checkpoint
- $mappings->write_to_file;
+ # split plugin name into module and method
+ $plugin =~ /(.*)::(\w+)$/;
+ my $module = $1;
+ my $method = $2;
- return $mappings;
-}
-
-
-sub higher_score_exists {
- my ($self, $entry, $matrix, $sources_done, $targets_done) = @_;
-
- my $source = $entry->source;
- my $target = $entry->target;
- my $score = $entry->score;
-
- foreach my $other_source (@{ $matrix->get_sources_for_target($target) }) {
- if ($other_source != $source and !$sources_done->{$other_source} and
- $score < $matrix->get_score($other_source, $target)) {
- return 1;
- }
+ unless ($module and $method) {
+ throw("Unable to determine module and method name from $plugin.\n");
}
- foreach my $other_target (@{ $matrix->get_targets_for_source($source) }) {
- if ($other_target != $target and !$targets_done->{$other_target} and
- $score < $matrix->get_score($source, $other_target)) {
- return 1;
- }
- }
-
- return 0;
-}
-
-
-#
-# find ambiguous mappings (see scores_similar() for definition)
-#
-sub ambiguous_mapping {
- my ($self, $entry, $matrix, $other_sources, $other_targets) = @_;
-
- my $source = $entry->source;
- my $target = $entry->target;
- my $score = $entry->score;
-
- my $retval = 0;
-
- foreach my $other_source (@{ $matrix->get_sources_for_target($target) }) {
- my $other_score = $matrix->get_score($other_source, $target);
+ # instantiate the plugin unless we already have an instance
+ my $plugin_instance;
+ if ($self->has_plugin($module)) {
- if ($other_source != $source and
- ($self->scores_similar($score, $other_score) or $score < $other_score)) {
- $retval = 1;
- push @{ $other_sources }, $other_source;
- }
- }
-
- foreach my $other_target (@{ $matrix->get_targets_for_source($source) }) {
- my $other_score = $matrix->get_score($source, $other_target);
-
- if ($other_target != $target and
- ($self->scores_similar($score, $other_score) or $score < $other_score)) {
- $retval = 1;
- push @{ $other_targets }, $other_target;
- }
- }
-
- return $retval;
-}
-
-
-#
-# rule for similarity taken from java code...
-#
-sub scores_similar {
- my ($self, $s1, $s2) = @_;
-
- # always give priority to exact matches over very similar ones
- return 0 if ($s1 == 1 and $s2 < 1);
-
- my $diff = $s1 -$s2;
- $diff = -$diff if ($diff < 0);
+ # re-use an existing plugin instance
+ $plugin_instance = $self->get_plugin($module);
- my $pc = 2 * $diff / ($s1 + $s2);
-
- return ($pc < SIMILAR_SCORE_RATIO);
-}
-
-
-sub filter_sources {
- my ($self, $other_sources, $sources_done) = @_;
-
- unless (scalar(@$other_sources) and scalar(keys %$sources_done)) {
- return $other_sources;
- }
-
- my @tmp = ();
+ } else {
+
+ # inject and instantiate the plugin module
+ inject($module);
+ $plugin_instance = $module->new(
+ -LOGGER => $self->logger,
+ -CONF => $self->conf,
+ -CACHE => $self->cache
+ );
+ $self->add_plugin($plugin_instance);
- foreach my $e (@{ $other_sources }) {
- push @tmp, $e unless ($sources_done->{$e});
}
- return \@tmp;
+ # run the method on the plugin
+ #
+ # pass in a sequence number (number of method run, used for generating
+ # checkpoint files), the scores used for determining the mapping, and all
+ # other arguments passed to this method (these will vary for different object
+ # types)
+ #
+ # return the scores and mappings to feed into the next plugin in the chain
+ return $plugin_instance->$method($num, $score_builder, $mappings, $scores, @_);
}
-sub filter_targets {
- my ($self, $other_targets, $targets_done) = @_;
-
- unless (scalar(@{ $other_targets }) and scalar(keys %$targets_done)) {
- return $other_targets;
- }
-
- my @tmp = ();
-
- foreach my $e (@{ $other_targets }) {
- push @tmp, $e unless ($targets_done->{$e});
- }
+sub has_plugin {
+ my $self = shift;
+ my $module = shift;
- return \@tmp;
+ defined($self->{'_plugins'}->{$module}) ? (return 1) : (return 0);
}
-#
-# modified basic mapper that maps transcripts that are ambiguous within one gene
-#
-sub same_gene_transcript_mapping {
+sub get_plugin {
my $self = shift;
- my $matrix = shift;
- my $mapping_name = shift;
-
- # argument checks
- unless ($matrix and
- $matrix->isa('Bio::EnsEMBL::IdMapping::ScoredMappingMatrix')) {
- throw('Need a Bio::EnsEMBL::IdMapping::ScoredMappingMatrix.');
- }
-
- throw('Need a name for serialising the mapping.') unless ($mapping_name);
-
- # Create a new MappingList object. Specify AUTO_LOAD to load serialised
- # existing mappings if found
- my $dump_path = path_append($self->conf->param('basedir'), 'mapping');
-
- my $mappings = Bio::EnsEMBL::IdMapping::MappingList->new(
- -DUMP_PATH => $dump_path,
- -CACHE_FILE => "${mapping_name}.ser",
- -AUTO_LOAD => 1,
- );
-
- # checkpoint test: return a previously stored MappingList
- if ($mappings->loaded) {
- $self->logger->info("Read existing mappings from ${mapping_name}.ser.\n");
- return $mappings;
- }
-
- my $sources_done = {};
- my $targets_done = {};
-
- # sort scoring matrix entries by descending score
- my @sorted_entries = sort { $b->score <=> $a->score ||
- $a->source <=> $b->source || $a->target <=> $b->target }
- @{ $matrix->get_all_Entries };
-
- while (my $entry = shift(@sorted_entries)) {
-
- # $self->logger->debug("\nxxx4 ".$entry->to_string." ");
-
- # we already found a mapping for either source or target yet
- next if ($sources_done->{$entry->source} or
- $targets_done->{$entry->target});
-
- #$self->logger->debug('d');
-
- my $other_sources = [];
- my $other_targets = [];
- my %source_genes = ();
- my %target_genes = ();
-
- if ($self->ambiguous_mapping($entry, $matrix, $other_sources, $other_targets)) {
- #$self->logger->debug('a');
-
- $other_sources = $self->filter_sources($other_sources, $sources_done);
- $other_targets = $self->filter_targets($other_targets, $targets_done);
+ my $module = shift;
- $source_genes{$self->cache->get_by_key('genes_by_transcript_id',
- 'source', $entry->source)} = 1;
- $target_genes{$self->cache->get_by_key('genes_by_transcript_id',
- 'target', $entry->target)} = 1;
-
- foreach my $other_source (@{ $other_sources }) {
- $source_genes{$self->cache->get_by_key('genes_by_transcript_id',
- 'source', $other_source)} = 1;
- }
-
- foreach my $other_target (@{ $other_targets }) {
- $target_genes{$self->cache->get_by_key('genes_by_transcript_id',
- 'target', $other_target)} = 1;
- }
-
- # only add mapping if only one source and target gene involved
- if (scalar(keys %source_genes) == 1 and scalar(keys %target_genes) == 1) {
- #$self->logger->debug('O');
- $mappings->add_Entry($entry);
- }
-
- } else {
- #$self->logger->debug('A');
-
- # this is the best mapping, add it
- $mappings->add_Entry($entry);
- }
+ return $self->{'_plugins'}->{$module};
+}
- $sources_done->{$entry->source} = 1;
- $targets_done->{$entry->target} = 1;
- }
- # create checkpoint
- $mappings->write_to_file;
+sub add_plugin {
+ my $self = shift;
+ my $plugin_instance = shift;
- return $mappings;
+ $self->{'_plugins'}->{ref($plugin_instance)} = $plugin_instance;
}
diff --git a/modules/Bio/EnsEMBL/IdMapping/InternalIdMapper/BaseMapper.pm b/modules/Bio/EnsEMBL/IdMapping/InternalIdMapper/BaseMapper.pm
new file mode 100644
index 0000000000..77a6c8208f
--- /dev/null
+++ b/modules/Bio/EnsEMBL/IdMapping/InternalIdMapper/BaseMapper.pm
@@ -0,0 +1,250 @@
+package Bio::EnsEMBL::IdMapping::InternalIdMapper::BaseMapper;
+
+=head1 NAME
+
+
+=head1 SYNOPSIS
+
+
+=head1 DESCRIPTION
+
+
+=head1 METHODS
+
+
+=head1 LICENCE
+
+This code is distributed under an Apache style licence. Please see
+http:#www.ensembl.org/info/about/code_licence.html for details.
+
+=head1 AUTHOR
+
+Patrick Meidl <meidl@ebi.ac.uk>, Ensembl core API team
+
+=head1 CONTACT
+
+Please post comments/questions to the Ensembl development list
+<ensembl-dev@ebi.ac.uk>
+
+=cut
+
+
+use strict;
+use warnings;
+no warnings 'uninitialized';
+
+use Bio::EnsEMBL::IdMapping::BaseObject;
+our @ISA = qw(Bio::EnsEMBL::IdMapping::BaseObject);
+
+use Bio::EnsEMBL::Utils::Exception qw(throw warning);
+use Bio::EnsEMBL::Utils::ScriptUtils qw(path_append);
+use Bio::EnsEMBL::IdMapping::MappingList;
+
+
+# scores are considered the same if (2.0 * (s1-s2))/(s1 + s2) < this
+use constant SIMILAR_SCORE_RATIO => 0.01;
+
+
+#
+# find the highest unambiguous score for all sources and targets in a scoring
+# matrix
+#
+sub basic_mapping {
+ my $self = shift;
+ my $matrix = shift;
+ my $mapping_name = shift;
+
+ # argument checks
+ unless ($matrix and
+ $matrix->isa('Bio::EnsEMBL::IdMapping::ScoredMappingMatrix')) {
+ throw('Need a Bio::EnsEMBL::IdMapping::ScoredMappingMatrix.');
+ }
+
+ throw('Need a name for serialising the mapping.') unless ($mapping_name);
+
+ # Create a new MappingList object. Specify AUTO_LOAD to load serialised
+ # existing mappings if found
+ my $dump_path = path_append($self->conf->param('basedir'), 'mapping');
+
+ my $mappings = Bio::EnsEMBL::IdMapping::MappingList->new(
+ -DUMP_PATH => $dump_path,
+ -CACHE_FILE => "${mapping_name}.ser",
+ -AUTO_LOAD => 1,
+ );
+
+ # checkpoint test: return a previously stored MappingList
+ if ($mappings->loaded) {
+ $self->logger->info("Read existing mappings from ${mapping_name}.ser.\n");
+ return $mappings;
+ }
+
+ my $sources_done = {};
+ my $targets_done = {};
+
+ # sort scoring matrix entries by descending score
+ my @sorted_entries = sort { $b->score <=> $a->score }
+ @{ $matrix->get_all_Entries };
+
+ # debug
+ #my $idx = substr($mapping_name, -1);
+
+ while (my $entry = shift(@sorted_entries)) {
+
+ #$self->logger->debug("\nxxx$idx ".$entry->to_string." ");
+
+ # we already found a mapping for either source or target
+ next if ($sources_done->{$entry->source} or
+ $targets_done->{$entry->target});
+
+ #$self->logger->debug('d');
+
+ # there's a better mapping for either source or target
+ next if ($self->higher_score_exists($entry, $matrix, $sources_done,
+ $targets_done));
+
+ #$self->logger->debug('h');
+
+ # check for ambiguous mappings; they are dealt with later
+ my $other_sources = [];
+ my $other_targets = [];
+
+ if ($self->ambiguous_mapping($entry, $matrix, $other_sources, $other_targets)) {
+ #$self->logger->debug('a');
+
+ $other_sources = $self->filter_sources($other_sources, $sources_done);
+ $other_targets = $self->filter_targets($other_targets, $targets_done);
+
+ next if (scalar(@$other_sources) or scalar(@$other_targets));
+ }
+
+ #$self->logger->debug('A');
+
+ # this is the best mapping, add it
+ $mappings->add_Entry($entry);
+
+ $sources_done->{$entry->source} = 1;
+ $targets_done->{$entry->target} = 1;
+ }
+
+ # create checkpoint
+ $mappings->write_to_file;
+
+ return $mappings;
+}
+
+
+sub higher_score_exists {
+ my ($self, $entry, $matrix, $sources_done, $targets_done) = @_;
+
+ my $source = $entry->source;
+ my $target = $entry->target;
+ my $score = $entry->score;
+
+ foreach my $other_source (@{ $matrix->get_sources_for_target($target) }) {
+ if ($other_source != $source and !$sources_done->{$other_source} and
+ $score < $matrix->get_score($other_source, $target)) {
+ return 1;
+ }
+ }
+
+ foreach my $other_target (@{ $matrix->get_targets_for_source($source) }) {
+ if ($other_target != $target and !$targets_done->{$other_target} and
+ $score < $matrix->get_score($source, $other_target)) {
+ return 1;
+ }
+ }
+
+ return 0;
+}
+
+
+#
+# find ambiguous mappings (see scores_similar() for definition)
+#
+sub ambiguous_mapping {
+ my ($self, $entry, $matrix, $other_sources, $other_targets) = @_;
+
+ my $source = $entry->source;
+ my $target = $entry->target;
+ my $score = $entry->score;
+
+ my $retval = 0;
+
+ foreach my $other_source (@{ $matrix->get_sources_for_target($target) }) {
+ my $other_score = $matrix->get_score($other_source, $target);
+
+ if ($other_source != $source and
+ ($self->scores_similar($score, $other_score) or $score < $other_score)) {
+ $retval = 1;
+ push @{ $other_sources }, $other_source;
+ }
+ }
+
+ foreach my $other_target (@{ $matrix->get_targets_for_source($source) }) {
+ my $other_score = $matrix->get_score($source, $other_target);
+
+ if ($other_target != $target and
+ ($self->scores_similar($score, $other_score) or $score < $other_score)) {
+ $retval = 1;
+ push @{ $other_targets }, $other_target;
+ }
+ }
+
+ return $retval;
+}
+
+
+#
+# rule for similarity taken from java code...
+#
+sub scores_similar {
+ my ($self, $s1, $s2) = @_;
+
+ # always give priority to exact matches over very similar ones
+ return 0 if ($s1 == 1 and $s2 < 1);
+
+ my $diff = $s1 -$s2;
+ $diff = -$diff if ($diff < 0);
+
+ my $pc = 2 * $diff / ($s1 + $s2);
+
+ return ($pc < SIMILAR_SCORE_RATIO);
+}
+
+
+sub filter_sources {
+ my ($self, $other_sources, $sources_done) = @_;
+
+ unless (scalar(@$other_sources) and scalar(keys %$sources_done)) {
+ return $other_sources;
+ }
+
+ my @tmp = ();
+
+ foreach my $e (@{ $other_sources }) {
+ push @tmp, $e unless ($sources_done->{$e});
+ }
+
+ return \@tmp;
+}
+
+
+sub filter_targets {
+ my ($self, $other_targets, $targets_done) = @_;
+
+ unless (scalar(@{ $other_targets }) and scalar(keys %$targets_done)) {
+ return $other_targets;
+ }
+
+ my @tmp = ();
+
+ foreach my $e (@{ $other_targets }) {
+ push @tmp, $e unless ($targets_done->{$e});
+ }
+
+ return \@tmp;
+}
+
+
+1;
+
diff --git a/modules/Bio/EnsEMBL/IdMapping/InternalIdMapper/EnsemblExonGeneric.pm b/modules/Bio/EnsEMBL/IdMapping/InternalIdMapper/EnsemblExonGeneric.pm
new file mode 100644
index 0000000000..7caae11a37
--- /dev/null
+++ b/modules/Bio/EnsEMBL/IdMapping/InternalIdMapper/EnsemblExonGeneric.pm
@@ -0,0 +1,91 @@
+package Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblExonGeneric;
+
+=head1 NAME
+
+
+=head1 SYNOPSIS
+
+
+=head1 DESCRIPTION
+
+
+=head1 METHODS
+
+
+=head1 LICENCE
+
+This code is distributed under an Apache style licence. Please see
+http:#www.ensembl.org/info/about/code_licence.html for details.
+
+=head1 AUTHOR
+
+Patrick Meidl <meidl@ebi.ac.uk>, Ensembl core API team
+
+=head1 CONTACT
+
+Please post comments/questions to the Ensembl development list
+<ensembl-dev@ebi.ac.uk>
+
+=cut
+
+
+use strict;
+use warnings;
+no warnings 'uninitialized';
+
+use Bio::EnsEMBL::IdMapping::InternalIdMapper::BaseMapper;
+our @ISA = qw(Bio::EnsEMBL::IdMapping::InternalIdMapper::BaseMapper);
+
+use Bio::EnsEMBL::Utils::Exception qw(throw warning);
+
+
+#
+# basic mapping
+#
+sub init_basic {
+ my $self = shift;
+ my $num = shift;
+ my $esb = shift;
+ my $mappings = shift;
+ my $exon_scores = shift;
+
+ $self->logger->info("Basic exon mapping...\n", 0, 'stamped');
+
+ $mappings = $self->basic_mapping($exon_scores, "exon_mappings$num");
+ $num++;
+ my $new_scores = $esb->create_shrinked_matrix($exon_scores, $mappings,
+ "exon_matrix$num");
+
+ return ($new_scores, $mappings);
+}
+
+
+#
+# reduce score for mappings of exons which do not belong to mapped
+# transcripts
+#
+sub mapped_transcript {
+ my $self = shift;
+ my $num = shift;
+ my $esb = shift;
+ my $mappings = shift;
+ my $exon_scores = shift;
+
+ $self->logger->info("Exons in mapped transcript...\n", 0, 'stamped');
+
+ unless ($exon_scores->loaded) {
+ $esb->non_mapped_transcript_rescore($exon_scores, $mappings);
+ $exon_scores->write_to_file;
+ }
+
+ $mappings = $self->basic_mapping($exon_scores, "exon_mappings$num");
+ $num++;
+ my $new_scores = $esb->create_shrinked_matrix($exon_scores, $mappings,
+ "exon_matrix$num");
+
+ return ($new_scores, $mappings);
+}
+
+
+1;
+
diff --git a/modules/Bio/EnsEMBL/IdMapping/InternalIdMapper/EnsemblGeneGeneric.pm b/modules/Bio/EnsEMBL/IdMapping/InternalIdMapper/EnsemblGeneGeneric.pm
new file mode 100644
index 0000000000..c424896f12
--- /dev/null
+++ b/modules/Bio/EnsEMBL/IdMapping/InternalIdMapper/EnsemblGeneGeneric.pm
@@ -0,0 +1,187 @@
+package Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblGeneGeneric;
+
+=head1 NAME
+
+Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblGeneGeneric - default Ensembl
+InternalIdMapper implementation for genes
+
+=head1 SYNOPSIS
+
+
+=head1 DESCRIPTION
+
+
+=head1 METHODS
+
+
+=head1 LICENCE
+
+This code is distributed under an Apache style licence. Please see
+http://www.ensembl.org/info/about/code_licence.html for details.
+
+=head1 AUTHOR
+
+Patrick Meidl <meidl@ebi.ac.uk>, Ensembl core API team
+
+=head1 CONTACT
+
+Please post comments/questions to the Ensembl development list
+<ensembl-dev@ebi.ac.uk>
+
+=cut
+
+
+use strict;
+use warnings;
+no warnings 'uninitialized';
+
+use Bio::EnsEMBL::IdMapping::InternalIdMapper::BaseMapper;
+our @ISA = qw(Bio::EnsEMBL::IdMapping::InternalIdMapper::BaseMapper);
+
+use Bio::EnsEMBL::Utils::Exception qw(throw warning);
+use Bio::EnsEMBL::Utils::ScriptUtils qw(path_append);
+
+
+#
+# basic mapping
+#
+sub init_basic {
+ my $self = shift;
+ my $num = shift;
+ my $gsb = shift;
+ my $mappings = shift;
+ my $gene_scores = shift;
+
+ $self->logger->info("Basic gene mapping...\n", 0, 'stamped');
+
+ $mappings = $self->basic_mapping($gene_scores, "gene_mappings$num");
+ $num++;
+
+ my $new_scores = $gsb->create_shrinked_matrix($gene_scores, $mappings,
+ "gene_matrix$num");
+
+ return ($new_scores, $mappings);
+}
+
+
+#
+# build the synteny from unambiguous mappings
+#
+sub synteny {
+ my $self = shift;
+ my $num = shift;
+ my $gsb = shift;
+ my $mappings = shift;
+ my $gene_scores = shift;
+
+ unless ($gene_scores->loaded) {
+ $self->logger->info("Synteny Framework building...\n", 0, 'stamped');
+ my $dump_path = path_append($self->conf->param('basedir'), 'mapping');
+ my $sf = Bio::EnsEMBL::IdMapping::SyntenyFramework->new(
+ -DUMP_PATH => $dump_path,
+ -CACHE_FILE => 'synteny_framework.ser',
+ -LOGGER => $self->logger,
+ -CONF => $self->conf,
+ -CACHE => $self->cache,
+ );
+ $sf->build_synteny($mappings);
+
+ # use it to rescore the genes
+ $self->logger->info("\nSynteny assisted mapping...\n", 0, 'stamped');
+ $gene_scores = $sf->rescore_gene_matrix_lsf($gene_scores);
+
+ # checkpoint
+ $gene_scores->write_to_file;
+ }
+
+ my $new_mappings = $self->basic_mapping($gene_scores, "gene_mappings$num");
+ $num++;
+ my $new_scores = $gsb->create_shrinked_matrix($gene_scores, $new_mappings,
+ "gene_matrix$num");
+
+ return ($new_scores, $new_mappings);
+}
+
+
+#
+# rescore with simple scoring function and try again
+#
+sub best_transcript {
+ my $self = shift;
+ my $num = shift;
+ my $gsb = shift;
+ my $mappings = shift;
+ my $gene_scores = shift;
+ my $transcript_scores = shift;
+
+ $self->logger->info("Retry with simple best transcript score...\n", 0, 'stamped');
+
+ unless ($gene_scores->loaded) {
+ $gsb->simple_gene_rescore($gene_scores, $transcript_scores);
+ $gene_scores->write_to_file;
+ }
+
+ my $new_mappings = $self->basic_mapping($gene_scores, "gene_mappings$num");
+ $num++;
+ my $new_scores = $gsb->create_shrinked_matrix($gene_scores, $new_mappings,
+ "gene_matrix$num");
+
+ return ($new_scores, $new_mappings);
+}
+
+
+#
+# rescore by penalising scores between genes with different biotypes
+#
+sub biotype {
+ my $self = shift;
+ my $num = shift;
+ my $gsb = shift;
+ my $mappings = shift;
+ my $gene_scores = shift;
+
+ $self->logger->info("Retry with biotype disambiguation...\n", 0, 'stamped');
+
+ unless ($gene_scores->loaded) {
+ $gsb->biotype_gene_rescore($gene_scores);
+ $gene_scores->write_to_file;
+ }
+
+ my $new_mappings = $self->basic_mapping($gene_scores, "gene_mappings$num");
+ $num++;
+ my $new_scores = $gsb->create_shrinked_matrix($gene_scores, $new_mappings,
+ "gene_matrix$num");
+
+ return ($new_scores, $new_mappings);
+}
+
+
+#
+# selectively rescore by penalising scores between genes with different
+# internalIDs
+#
+sub internal_id {
+ my $self = shift;
+ my $num = shift;
+ my $gsb = shift;
+ my $mappings = shift;
+ my $gene_scores = shift;
+
+ $self->logger->info("Retry with internalID disambiguation...\n", 0, 'stamped');
+
+ unless ($gene_scores->loaded) {
+ $gsb->internal_id_rescore($gene_scores);
+ $gene_scores->write_to_file;
+ }
+
+ my $new_mappings = $self->basic_mapping($gene_scores, "gene_mappings$num");
+ $num++;
+ my $new_scores = $gsb->create_shrinked_matrix($gene_scores, $new_mappings,
+ "gene_matrix$num");
+
+ return ($new_scores, $new_mappings);
+}
+
+
+1;
+
diff --git a/modules/Bio/EnsEMBL/IdMapping/InternalIdMapper/EnsemblTranscriptGeneric.pm b/modules/Bio/EnsEMBL/IdMapping/InternalIdMapper/EnsemblTranscriptGeneric.pm
new file mode 100644
index 0000000000..8c7ad9ab33
--- /dev/null
+++ b/modules/Bio/EnsEMBL/IdMapping/InternalIdMapper/EnsemblTranscriptGeneric.pm
@@ -0,0 +1,276 @@
+package Bio::EnsEMBL::IdMapping::InternalIdMapper::EnsemblTranscriptGeneric;
+
+=head1 NAME
+
+
+=head1 SYNOPSIS
+
+
+=head1 DESCRIPTION
+
+
+=head1 METHODS
+
+
+=head1 LICENCE
+
+This code is distributed under an Apache style licence. Please see
+http:#www.ensembl.org/info/about/code_licence.html for details.
+
+=head1 AUTHOR
+
+Patrick Meidl <meidl@ebi.ac.uk>, Ensembl core API team
+
+=head1 CONTACT
+
+Please post comments/questions to the Ensembl development list
+<ensembl-dev@ebi.ac.uk>
+
+=cut
+
+
+use strict;
+use warnings;
+no warnings 'uninitialized';
+
+use Bio::EnsEMBL::IdMapping::InternalIdMapper::BaseMapper;
+our @ISA = qw(Bio::EnsEMBL::IdMapping::InternalIdMapper::BaseMapper);
+
+use Bio::EnsEMBL::Utils::Exception qw(throw warning);
+use Bio::EnsEMBL::Utils::ScriptUtils qw(path_append);
+
+
+#
+# basic mapping
+#
+sub init_basic {
+ my $self = shift;
+ my $num = shift;
+ my $tsb = shift;
+ my $mappings = shift;
+ my $transcript_scores = shift;
+
+ $self->logger->info("Basic transcript mapping...\n", 0, 'stamped');
+
+ $mappings = $self->basic_mapping($transcript_scores,
+ "transcript_mappings$num");
+ $num++;
+ my $new_scores = $tsb->create_shrinked_matrix($transcript_scores, $mappings,
+ "transcript_matrix$num");
+
+ return ($new_scores, $mappings);
+}
+
+
+#
+# handle cases with exact match but different translation
+#
+sub non_exact_translation {
+ my $self = shift;
+ my $num = shift;
+ my $tsb = shift;
+ my $mappings = shift;
+ my $transcript_scores = shift;
+
+ $self->logger->info("Exact Transcript non-exact Translation...\n", 0, 'stamped');
+
+ unless ($transcript_scores->loaded) {
+ $tsb->different_translation_rescore($transcript_scores);
+ $transcript_scores->write_to_file;
+ }
+
+ $mappings = $self->basic_mapping($transcript_scores,
+ "transcript_mappings$num");
+ $num++;
+ my $new_scores = $tsb->create_shrinked_matrix($transcript_scores, $mappings,
+ "transcript_matrix$num");
+
+ return ($new_scores, $mappings);
+}
+
+
+#
+# reduce score for mappings of transcripts which do not belong to mapped
+# genes
+#
+sub mapped_gene {
+ my $self = shift;
+ my $num = shift;
+ my $tsb = shift;
+ my $mappings = shift;
+ my $transcript_scores = shift;
+ my $gene_mappings = shift;
+
+ $self->logger->info("Transcripts in mapped genes...\n", 0, 'stamped');
+
+ unless ($transcript_scores->loaded) {
+ $tsb->non_mapped_gene_rescore($transcript_scores, $gene_mappings);
+ $transcript_scores->write_to_file;
+ }
+
+ $mappings = $self->basic_mapping($transcript_scores,
+ "transcript_mappings$num");
+ $num++;
+ my $new_scores = $tsb->create_shrinked_matrix($transcript_scores, $mappings,
+ "transcript_matrix$num");
+
+ return ($new_scores, $mappings);
+}
+
+
+#
+# selectively rescore by penalising scores between transcripts with
+# different internalIDs
+#
+sub internal_id {
+ my $self = shift;
+ my $num = shift;
+ my $tsb = shift;
+ my $mappings = shift;
+ my $transcript_scores = shift;
+
+ $self->logger->info("Retry with internalID disambiguation...\n", 0, 'stamped');
+
+ unless ($transcript_scores->loaded) {
+ $tsb->internal_id_rescore($transcript_scores);
+ $transcript_scores->write_to_file;
+ }
+
+ $mappings = $self->basic_mapping($transcript_scores,
+ "transcript_mappings$num");
+ $num++;
+ my $new_scores = $tsb->create_shrinked_matrix($transcript_scores, $mappings,
+ "transcript_matrix$num");
+
+ return ($new_scores, $mappings);
+}
+
+
+#
+# handle ambiguities between transcripts in single genes
+#
+sub single_gene {
+ my $self = shift;
+ my $num = shift;
+ my $tsb = shift;
+ my $mappings = shift;
+ my $transcript_scores = shift;
+
+ $self->logger->info("Transcripts in single genes...\n", 0, 'stamped');
+
+ unless ($transcript_scores->loaded) {
+ $transcript_scores->write_to_file;
+ }
+
+ $mappings = $self->same_gene_transcript_mapping($transcript_scores,
+ "transcript_mappings$num");
+ $num++;
+ my $new_scores = $tsb->create_shrinked_matrix($transcript_scores, $mappings,
+ "transcript_matrix$num");
+
+ return ($new_scores, $mappings);
+}
+
+
+#
+# modified basic mapper that maps transcripts that are ambiguous within one gene
+#
+sub same_gene_transcript_mapping {
+ my $self = shift;
+ my $matrix = shift;
+ my $mapping_name = shift;
+
+ # argument checks
+ unless ($matrix and
+ $matrix->isa('Bio::EnsEMBL::IdMapping::ScoredMappingMatrix')) {
+ throw('Need a Bio::EnsEMBL::IdMapping::ScoredMappingMatrix.');
+ }
+
+ throw('Need a name for serialising the mapping.') unless ($mapping_name);
+
+ # Create a new MappingList object. Specify AUTO_LOAD to load serialised
+ # existing mappings if found
+ my $dump_path = path_append($self->conf->param('basedir'), 'mapping');
+
+ my $mappings = Bio::EnsEMBL::IdMapping::MappingList->new(
+ -DUMP_PATH => $dump_path,
+ -CACHE_FILE => "${mapping_name}.ser",
+ -AUTO_LOAD => 1,
+ );
+
+ # checkpoint test: return a previously stored MappingList
+ if ($mappings->loaded) {
+ $self->logger->info("Read existing mappings from ${mapping_name}.ser.\n");
+ return $mappings;
+ }
+
+ my $sources_done = {};
+ my $targets_done = {};
+
+ # sort scoring matrix entries by descending score
+ my @sorted_entries = sort { $b->score <=> $a->score ||
+ $a->source <=> $b->source || $a->target <=> $b->target }
+ @{ $matrix->get_all_Entries };
+
+ while (my $entry = shift(@sorted_entries)) {
+
+ # $self->logger->debug("\nxxx4 ".$entry->to_string." ");
+
+ # we already found a mapping for either source or target yet
+ next if ($sources_done->{$entry->source} or
+ $targets_done->{$entry->target});
+
+ #$self->logger->debug('d');
+
+ my $other_sources = [];
+ my $other_targets = [];
+ my %source_genes = ();
+ my %target_genes = ();
+
+ if ($self->ambiguous_mapping($entry, $matrix, $other_sources, $other_targets)) {
+ #$self->logger->debug('a');
+
+ $other_sources = $self->filter_sources($other_sources, $sources_done);
+ $other_targets = $self->filter_targets($other_targets, $targets_done);
+
+ $source_genes{$self->cache->get_by_key('genes_by_transcript_id',
+ 'source', $entry->source)} = 1;
+ $target_genes{$self->cache->get_by_key('genes_by_transcript_id',
+ 'target', $entry->target)} = 1;
+
+ foreach my $other_source (@{ $other_sources }) {
+ $source_genes{$self->cache->get_by_key('genes_by_transcript_id',
+ 'source', $other_source)} = 1;
+ }
+
+ foreach my $other_target (@{ $other_targets }) {
+ $target_genes{$self->cache->get_by_key('genes_by_transcript_id',
+ 'target', $other_target)} = 1;
+ }
+
+ # only add mapping if only one source and target gene involved
+ if (scalar(keys %source_genes) == 1 and scalar(keys %target_genes) == 1) {
+ #$self->logger->debug('O');
+ $mappings->add_Entry($entry);
+ }
+
+ } else {
+ #$self->logger->debug('A');
+
+ # this is the best mapping, add it
+ $mappings->add_Entry($entry);
+ }
+
+ $sources_done->{$entry->source} = 1;
+ $targets_done->{$entry->target} = 1;
+ }
+
+ # create checkpoint
+ $mappings->write_to_file;
+
+ return $mappings;
+}
+
+
+1;
+
--
GitLab