JAL-1265 basic support for parsing database accessions from Stockholm

JAL-851 - guess the primary sequence accession based on RFAM or PFAM alignment accession code

diff --git a/src/jalview/io/StockholmFile.java b/src/jalview/io/StockholmFile.java
index ff2b4fa..02ab5c1 100644 (file)
--- a/src/jalview/io/StockholmFile.java
+++ b/src/jalview/io/StockholmFile.java
@@ -146,7 +146,7 @@ public class StockholmFile extends AlignFile
         // End of the alignment, pass stuff back
         this.noSeqs = seqs.size();
 
-        String propety = null;
+        String seqdb,dbsource = null;
         Regex pf = new Regex("PF[0-9]{5}(.*)"); // Finds AC for Pfam
         Regex rf = new Regex("RF[0-9]{5}(.*)"); // Finds AC for Rfam
         if (getAlignmentProperty("AC") != null)
@@ -154,11 +154,12 @@ public class StockholmFile extends AlignFile
           String dbType = getAlignmentProperty("AC").toString();
           if (pf.search(dbType))
           {
-            propety = "PFAM";
+            // PFAM Alignment - so references are typically from Uniprot
+            dbsource = "PFAM";
           }
           else if (rf.search(dbType))
           {
-            propety = "RFAM";
+            dbsource = "RFAM";
           }
         }
         // logger.debug("Number of sequences: " + this.noSeqs);
@@ -216,16 +217,19 @@ public class StockholmFile extends AlignFile
             }
           }
 
-          if (accAnnotations != null && accAnnotations.containsKey("AC")
-                  && propety != null)
+          if (accAnnotations != null && accAnnotations.containsKey("AC"))
           {
-            String dbr = (String) accAnnotations.get("AC");
-            if (dbr != null)
+            if (dbsource != null)
             {
-              String src = propety;
-              String acn = dbr.toString();
-              jalview.util.DBRefUtils.parseToDbRef(seqO, src, "0", acn);
-            }
+              String dbr = (String) accAnnotations.get("AC");
+              if (dbr != null)
+              {
+                // we could get very clever here - but for now - just try to guess accession type from source of alignment plus structure of accession
+                guessDatabaseFor(seqO, dbr, dbsource);
+                
+              }
+            } 
+            // else - do what ?  add the data anyway and prompt the user to specify what references these are ?
           }
 
           Hashtable features = null;
@@ -563,6 +567,95 @@ public class StockholmFile extends AlignFile
     }
   }
 
+  /**
+   * Demangle an accession string and guess the originating sequence database for a given sequence
+   * @param seqO sequence to be annotated
+   * @param dbr Accession string for sequence
+   * @param dbsource source database for alignment (PFAM or RFAM)
+   */
+  private void guessDatabaseFor(Sequence seqO, String dbr, String dbsource)
+  {
+    DBRefEntry dbrf=null;
+    List<DBRefEntry> dbrs=new ArrayList<DBRefEntry>();
+    String seqdb="Unknown",sdbac=""+dbr;
+    int st=-1,en=-1,p;
+    if ((st=sdbac.indexOf("/"))>-1)
+    {
+      String num,range=sdbac.substring(st+1);
+      sdbac = sdbac.substring(0,st);
+      if ((p=range.indexOf("-"))>-1)
+      {
+        p++;
+        if (p<range.length())
+        {
+        num = range.substring(p).trim();
+        try {
+          en = Integer.parseInt(num);
+        } catch (NumberFormatException x)
+        {
+          // could warn here that index is invalid
+          en = -1;
+        }
+        }
+      } else {
+        p=range.length();
+      }
+      num=range.substring(0,p).trim();
+      try {
+        st = Integer.parseInt(num);
+      } catch (NumberFormatException x)
+      {
+        // could warn here that index is invalid
+        st = -1;
+      }
+    }
+    if (dbsource.equals("PFAM")) {
+      seqdb = "UNIPROT";
+      if (sdbac.indexOf(".")>-1)
+      {
+        // strip of last subdomain
+        sdbac = sdbac.substring(0,sdbac.indexOf("."));
+        dbrf = jalview.util.DBRefUtils.parseToDbRef(seqO, seqdb, dbsource, sdbac);
+        if (dbrf!=null)
+        {
+          dbrs.add(dbrf);
+        }
+      }
+      dbrf = jalview.util.DBRefUtils.parseToDbRef(seqO, dbsource, dbsource, dbr);
+      if (dbr!=null)
+      {
+        dbrs.add(dbrf);
+      }
+    } else {
+      seqdb = "EMBL"; // total guess - could be ENA, or something else these days
+      if (sdbac.indexOf(".")>-1)
+      {
+        // strip off last subdomain
+        sdbac = sdbac.substring(0,sdbac.indexOf("."));
+        dbrf = jalview.util.DBRefUtils.parseToDbRef(seqO, seqdb, dbsource, sdbac);
+        if (dbrf!=null)
+        {
+          dbrs.add(dbrf);
+        }
+      }
+      
+      dbrf = jalview.util.DBRefUtils.parseToDbRef(seqO, dbsource, dbsource, dbr);
+      if (dbrf!=null)
+      {
+        dbrs.add(dbrf);
+      }
+    }
+    if (st!=-1 && en!=-1)
+    {
+      for (DBRefEntry d:dbrs)
+      {
+        jalview.util.MapList mp = new jalview.util.MapList(new int[] { seqO.getStart(),seqO.getEnd()},new int[] { st,en},1,1);
+        jalview.datamodel.Mapping mping = new Mapping(mp);
+        d.setMap(mping);
+      }
+    }
+  }
+
   protected static AlignmentAnnotation parseAnnotationRow(
           Vector annotation, String label, String annots)
   {