changes from branch-vega-31-dev

modules/Bio/EnsEMBL/Gene.pm

@@ -461,11 +461,40 @@ sub get_all_homologous_Genes{
 
 
 sub type {
-   my $self = shift;
-   
-   $self->{'biotype'} = shift if( @_ );
-
-   return ( $self->{'biotype'} || "protein_coding" );
+    my $self = shift;
+    
+    $self->{'biotype'} = shift if( @_ );
+    
+    # map biotype/confidence to HAWK classification for Vega
+    if ($self->source eq 'vega') {
+        my %typemap = (
+            'protein_coding.KNOWN'          => 'Known',
+            'protein_coding.NOVEL'          => 'Novel_CDS',
+            'unclassified.NOVEL'            => 'Novel_Transcript',
+            'protein_coding.PREDICTED'      => 'Predicted_Gene',
+            'unclassified.PUTATIVE'         => 'Putative',
+            'pseudogene.KNOWN'              => 'Pseudogene',
+            'pseudogene.NOVEL'              => 'Pseudogene',
+            'processed_pseudogene.KNOWN'    => 'Processed_pseudogene',
+            'processed_pseudogene.NOVEL'    => 'Processed_pseudogene',
+            'unprocessed_pseudogene.KNOWN'  => 'Unprocessed_pseudogene',
+            'unprocessed_pseudogene.NOVEL'  => 'Unprocessed_pseudogene',
+            'Ig_pseudogene_segment.KNOWN'   => 'Ig_Pseudogene_Segment',
+            'Ig_pseudogene_segment.NOVEL'   => 'Ig_Pseudogene_Segment',
+            'Ig_segment.KNOWN'              => 'Ig_Segment',
+            'Ig_segment.NOVEL'              => 'Ig_Segment',
+        );
+        my $bio_conf = $self->biotype.".".$self->confidence;
+        if ($typemap{$bio_conf}) {
+            return $typemap{$bio_conf};
+        } else {
+            warning("biotype.confidence ($bio_conf) cannot be resolved to HAWK type");
+        }
+
+    # for all other sources, return biotype
+    } else {
+        return ( $self->{'biotype'} || "protein_coding" );
+    }
 }
 
 

sql/patch_30_31.sql

@@ -21,14 +21,14 @@ ALTER TABLE transcript ADD description text;
 UPDATE transcript t, xref x SET t.description = x.description WHERE t.display_xref_id = x.xref_id;
 UPDATE transcript SET description=NULL WHERE description="";
 
-
 # usefull settings for the new tables
-UPDATE gene SET biotype='protein_coding' WHERE biotype = 'ensembl';
- 
-UPDATE gene SET confidence='KNOWN' WHERE display_xref_id != 0;
-UPDATE transcript SET confidence='KNOWN' WHERE display_xref_id != 0;
+
+UPDATE gene SET source = 'ensembl';
+UPDATE gene g, xref x, external_db ed SET g.confidence='KNOWN' WHERE g.display_xref_id = x.xref_id and x.external_db_id = ed.external_db_id and g.display_xref_id != 0 and ed.status like 'KNOWN%';
+UPDATE transcript t, xref x, external_db ed SET t.confidence='KNOWN' WHERE t.display_xref_id = x.xref_id and x.external_db_id = ed.external_db_id and t.display_xref_id != 0 and ed.status like 'KNOWN%';
 
 # some vega specific stuff, shouldnt harm anybody else
+
 UPDATE gene SET biotype='unclassified' WHERE biotype = 'Transcript';
 UPDATE gene SET biotype='pseudogene' WHERE biotype = 'Pseudogene';
 UPDATE gene SET biotype='protein_coding', confidence='NOVEL' WHERE biotype = 'Novel_CDS';
@@ -51,7 +51,6 @@ UPDATE transcript t, gene g SET t.biotype = g.biotype WHERE g.gene_id = t.gene_i
 
 # new tables regulatory stuff and transcript supporting features
 
-
 ################################################################################
 #
 # Table structure for table 'regulatory_feature'

sql/vega_latest_schema.sql

 ###################################################################
-#FROM patch_23_24.sql	
+# FROM patch_23_24.sql	
 
 # Adds a display_label column to the prediction_transcript table and
 # populates it.
@@ -42,3 +42,60 @@ WHERE  a.analysis_id = pt.analysis_id;
 
 DROP TABLE prediction_transcript;
 ALTER TABLE tmp_prediction_transcript RENAME prediction_transcript;
+
+
+
+###################################################################
+# FROM patch_30_31.sql
+
+
+# gene table changes
+
+ALTER TABLE gene CHANGE type biotype VARCHAR(40) NOT NULL default 'protein_coding';
+ALTER TABLE gene ADD source VARCHAR(20) NOT NULL default 'ensembl';
+ALTER TABLE gene ADD confidence ENUM( 'KNOWN', 'NOVEL', 'PUTATIVE', 'PREDICTED' ) default 'NOVEL';
+ALTER TABLE gene ADD description text;
+
+UPDATE gene g, gene_description gd SET g.description = gd.description WHERE gd.gene_id = g.gene_id;
+
+DROP TABLE gene_description;
+
+# transcript related changes
+
+ALTER TABLE transcript ADD biotype VARCHAR(40) NOT NULL DEFAULT 'protein_coding';
+ALTER TABLE transcript ADD confidence ENUM( 'KNOWN', 'NOVEL', 'PUTATIVE', 'PREDICTED' ) default 'NOVEL';
+ALTER TABLE transcript ADD description text;
+
+# reasonable default for transcript description
+# it might be questionable whether a separate transcript description is necessary
+ 
+UPDATE transcript t, xref x SET t.description = x.description WHERE t.display_xref_id = x.xref_id;
+UPDATE transcript SET description=NULL WHERE description="";
+
+# usefull settings for the new tables
+
+UPDATE gene SET source = 'vega';
+UPDATE gene g, xref x, external_db ed SET g.confidence='KNOWN' WHERE g.display_xref_id = x.xref_id and x.external_db_id = ed.external_db_id and g.display_xref_id != 0 and ed.status like 'KNOWN%';
+UPDATE transcript t, xref x, external_db ed SET t.confidence='KNOWN' WHERE t.display_xref_id = x.xref_id and x.external_db_id = ed.external_db_id and t.display_xref_id != 0 and ed.status like 'KNOWN%';
+
+# some vega specific stuff, shouldnt harm anybody else
+
+UPDATE gene SET biotype='unclassified' WHERE biotype = 'Transcript';
+UPDATE gene SET biotype='pseudogene' WHERE biotype = 'Pseudogene';
+UPDATE gene SET biotype='protein_coding', confidence='NOVEL' WHERE biotype = 'Novel_CDS';
+UPDATE gene SET biotype='unclassified', confidence='NOVEL' WHERE biotype = 'Novel_Transcript';
+UPDATE gene SET biotype='unclassified',confidence='PUTATIVE' WHERE biotype = 'Putative';
+UPDATE gene SET biotype='protein_coding', confidence='KNOWN' WHERE biotype = 'Known';
+
+UPDATE gene SET biotype='processed_pseudogene' WHERE biotype = 'Processed_pseudogene';
+UPDATE gene SET biotype='unprocessed_pseudogene' WHERE biotype = 'Unprocessed_pseudogene';
+UPDATE gene SET biotype='protein_coding',confidence='PREDICTED' WHERE biotype = 'Predicted_Gene';
+UPDATE gene SET biotype='Ig_segment' WHERE biotype = 'Ig_Segment';
+UPDATE gene SET biotype='Ig_pseudogene_segment' WHERE biotype = 'Ig_Pseudogene_Segment';
+
+UPDATE gene SET biotype=replace( biotype, '-','_' );
+
+# reasonable biotypes for the transcripts, take the one from the gene
+
+UPDATE transcript t, gene g SET t.biotype = g.biotype WHERE g.gene_id = t.gene_id;
+