in progress

[jalview.git] / forester / java / src / org / forester / ws / seqdb / UniProtEntry.java

diff --git a/forester/java/src/org/forester/ws/seqdb/UniProtEntry.java b/forester/java/src/org/forester/ws/seqdb/UniProtEntry.java
index 029c4b1..3f0df32 100644 (file)
--- a/forester/java/src/org/forester/ws/seqdb/UniProtEntry.java
+++ b/forester/java/src/org/forester/ws/seqdb/UniProtEntry.java
@@ -36,6 +36,8 @@ import org.forester.go.GoNameSpace;
 import org.forester.go.GoTerm;
 import org.forester.phylogeny.data.Accession;
 import org.forester.phylogeny.data.Annotation;
+import org.forester.sequence.BasicSequence;
+import org.forester.sequence.MolecularSequence;
 import org.forester.util.ForesterUtil;
 
 public final class UniProtEntry implements SequenceDatabaseEntry {
@@ -51,6 +53,9 @@ public final class UniProtEntry implements SequenceDatabaseEntry {
     public final static Pattern  PDB_PATTERN       = Pattern.compile( "PDB;\\s+([0-9A-Z]{4});\\s+([^;]+)" );
     public final static Pattern  PharmGKB_PATTERN  = Pattern.compile( "PharmGKB;\\s+([0-9A-Z]+);" );
     public final static Pattern  Reactome_PATTERN  = Pattern.compile( "Reactome;\\s+([0-9A-Z]+);\\s+([^\\.]+)" );
+    public final static Pattern  HGNC_PATTERN      = Pattern.compile( "HGNC;\\s+HGNC:(\\d+);" );
+    public final static Pattern  NCBI_TAXID_PATTERN= Pattern.compile( "NCBI_TaxID=(\\d+)" );
+    
     private String               _ac;
     private SortedSet<Accession> _cross_references;
     private String               _gene_name;
@@ -59,6 +64,7 @@ public final class UniProtEntry implements SequenceDatabaseEntry {
     private String               _os_scientific_name;
     private String               _symbol;
     private String               _tax_id;
+    private MolecularSequence    _mol_seq;
 
     private UniProtEntry() {
     }
@@ -119,7 +125,7 @@ public final class UniProtEntry implements SequenceDatabaseEntry {
                 && ForesterUtil.isEmpty( getTaxonomyScientificName() ) && ForesterUtil.isEmpty( getSequenceSymbol() )
                 && ForesterUtil.isEmpty( getGeneName() ) && ForesterUtil.isEmpty( getTaxonomyIdentifier() )
                 && ForesterUtil.isEmpty( getSequenceSymbol() ) && ( ( getGoTerms() == null ) || getGoTerms().isEmpty() ) && ( ( getCrossReferences() == null ) || getCrossReferences()
-                .isEmpty() ) );
+                        .isEmpty() ) );
     }
 
     private void addCrossReference( final Accession accession ) {
@@ -142,6 +148,10 @@ public final class UniProtEntry implements SequenceDatabaseEntry {
         }
     }
 
+    private void setMolecularSequence( final MolecularSequence mol_seq ) {
+        _mol_seq = mol_seq;
+    }
+
     private void setGeneName( final String gene_name ) {
         if ( _gene_name == null ) {
             _gene_name = gene_name;
@@ -172,6 +182,9 @@ public final class UniProtEntry implements SequenceDatabaseEntry {
 
     public static SequenceDatabaseEntry createInstanceFromPlainText( final List<String> lines ) {
         final UniProtEntry e = new UniProtEntry();
+        boolean saw_sq = false;
+        final StringBuffer sq_buffer = new StringBuffer();
+        boolean is_aa = false;
         for( final String line : lines ) {
             //System.out.println( line );
             if ( line.startsWith( "AC" ) ) {
@@ -179,20 +192,43 @@ public final class UniProtEntry implements SequenceDatabaseEntry {
             }
             else if ( line.startsWith( "DE" ) && ForesterUtil.isEmpty( e.getSequenceName() ) ) {
                 if ( ( line.indexOf( "RecName:" ) > 0 ) && ( line.indexOf( "Full=" ) > 0 ) ) {
-                    e.setSequenceName( SequenceDbWsTools.extractFromTo( line, "Full=", ";" ) );
+                    if ( line.indexOf( "{" ) > 0  ) {
+                        e.setSequenceName( SequenceDbWsTools.extractFromTo( line, "Full=", "{" ) );
+                    }
+                    else {
+                        e.setSequenceName( SequenceDbWsTools.extractFromTo( line, "Full=", ";" ) );
+                    }
                 }
                 else if ( ( line.indexOf( "SubName:" ) > 0 ) && ( line.indexOf( "Full=" ) > 0 ) ) {
-                    e.setSequenceName( SequenceDbWsTools.extractFromTo( line, "Full=", ";" ) );
+                    if ( line.indexOf( "{" ) > 0  ) {
+                        e.setSequenceName( SequenceDbWsTools.extractFromTo( line, "Full=", "{" ) );
+                    }
+                    else {
+                        e.setSequenceName( SequenceDbWsTools.extractFromTo( line, "Full=", ";" ) );
+                    }
+                    
                 }
             }
             else if ( line.startsWith( "DE" ) && ForesterUtil.isEmpty( e.getSequenceSymbol() ) ) {
+              
                 if ( line.indexOf( "Short=" ) > 0 ) {
-                    e.setSequenceSymbol( SequenceDbWsTools.extractFromTo( line, "Short=", ";" ) );
+                    if ( line.indexOf( "{" ) > 0  ) {
+                        e.setSequenceSymbol( SequenceDbWsTools.extractFromTo( line, "Short=", "{" ) );
+                    }
+                    else {
+                        e.setSequenceSymbol( SequenceDbWsTools.extractFromTo( line, "Short=", ";" ) );
+                    }
+                   
                 }
             }
             else if ( line.startsWith( "GN" ) && ForesterUtil.isEmpty( e.getGeneName() ) ) {
                 if ( line.indexOf( "Name=" ) > 0 ) {
-                    e.setGeneName( SequenceDbWsTools.extractFromTo( line, "Name=", ";" ) );
+                    if ( line.indexOf( "{" ) > 0  ) {
+                        e.setGeneName( SequenceDbWsTools.extractFromTo( line, "Name=", "{" ) );
+                    }
+                    else {
+                        e.setGeneName( SequenceDbWsTools.extractFromTo( line, "Name=", ";" ) );
+                    }
                 }
             }
             else if ( line.startsWith( "DR" ) ) {
@@ -272,6 +308,12 @@ public final class UniProtEntry implements SequenceDatabaseEntry {
                         e.addCrossReference( new Accession( m.group( 1 ), "Reactome", m.group( 2 ) ) );
                     }
                 }
+                else if ( line.indexOf( "HGNC;" ) > 0 ) {
+                    final Matcher m = HGNC_PATTERN.matcher( line );
+                    if ( m.find() ) {
+                        e.addCrossReference( new Accession( m.group( 1 ), "HGNC" ) );
+                    }
+                }
             }
             else if ( line.startsWith( "OS" ) ) {
                 if ( line.indexOf( "(" ) > 0 ) {
@@ -283,9 +325,24 @@ public final class UniProtEntry implements SequenceDatabaseEntry {
             }
             else if ( line.startsWith( "OX" ) ) {
                 if ( line.indexOf( "NCBI_TaxID=" ) > 0 ) {
-                    e.setTaxId( SequenceDbWsTools.extractFromTo( line, "NCBI_TaxID=", ";" ) );
+                    final Matcher m = NCBI_TAXID_PATTERN.matcher( line );
+                    if ( m.find() ) {
+                        e.setTaxId( m.group( 1 ) );
+                    }
                 }
             }
+            else if ( line.startsWith( "SQ" ) ) {
+                saw_sq = true;
+                if ( line.contains( "AA;" ) ) {
+                    is_aa = true;
+                }
+            }
+            else if ( saw_sq && line.startsWith( " " ) ) {
+                sq_buffer.append( line.replaceAll( "\\s+", "" ) );
+            }
+        }
+        if ( ( sq_buffer.length() > 0 ) && is_aa ) {
+            e.setMolecularSequence( BasicSequence.createAaSequence( e.getAccession(), sq_buffer.toString() ) );
         }
         return e;
     }
@@ -294,4 +351,19 @@ public final class UniProtEntry implements SequenceDatabaseEntry {
     public SortedSet<Annotation> getAnnotations() {
         return null;
     }
+
+    @Override
+    public String getMap() {
+        return null;
+    }
+
+    @Override
+    public String getChromosome() {
+        return null;
+    }
+
+    @Override
+    public MolecularSequence getMolecularSequence() {
+        return _mol_seq;
+    }
 }