JAL-3692 unit test (J03321), fixes, dbrefs; todo: protein mappings

[jalview.git] / test / jalview / io / EmblFlatFileTest.java

diff --git a/test/jalview/io/EmblFlatFileTest.java b/test/jalview/io/EmblFlatFileTest.java
new file mode 100644 (file)
index 0000000..6d9874e
--- /dev/null
+++ b/test/jalview/io/EmblFlatFileTest.java
@@ -0,0 +1,136 @@
+package jalview.io;
+
+import static org.testng.Assert.assertEquals;
+import static org.testng.Assert.assertTrue;
+import static org.testng.Assert.assertNull;
+
+import java.io.File;
+import java.io.IOException;
+import java.net.MalformedURLException;
+import java.util.List;
+import java.util.Set;
+
+import org.testng.annotations.Test;
+
+import jalview.datamodel.DBRefEntry;
+import jalview.datamodel.SequenceFeature;
+import jalview.datamodel.SequenceI;
+import jalview.datamodel.features.SequenceFeatures;
+
+public class EmblFlatFileTest
+{
+  /**
+   * A fairly tough test, using J03321 (circular DNA), which has 8 CDS features,
+   * one of them reverse strand
+   * 
+   * @throws MalformedURLException
+   * @throws IOException
+   */
+  @Test(groups = "Functional")
+  public void testParse() throws MalformedURLException, IOException
+  {
+    File dataFile = new File("test/jalview/io/J03321.embl.txt");
+    FileParse fp = new FileParse(dataFile, DataSourceType.FILE);
+    EmblFlatFile parser = new EmblFlatFile(fp, "EmblTest");
+    parser.parse();
+    List<SequenceI> seqs = parser.getSeqs();
+
+    assertEquals(seqs.size(), 1);
+    SequenceI seq = seqs.get(0);
+    assertEquals(seq.getName(), "EmblTest|J03321");
+    assertEquals(seq.getLength(), 7502);
+
+    /*
+     * should be 9 CDS features (one is a 'join' of two exons)
+     */
+    Set<String> featureTypes = seq.getFeatures().getFeatureTypes();
+    assertEquals(featureTypes.size(), 1);
+    assertTrue(featureTypes.contains("CDS"));
+    
+    /*
+     * inspect some features (sort them for convenience of test assertions)
+     */
+    List<SequenceFeature> features = seq.getFeatures()
+            .getAllFeatures("CDS");
+    SequenceFeatures.sortFeatures(features,  true);
+    assertEquals(features.size(), 9);
+    
+    SequenceFeature sf = features.get(0);
+    assertEquals(sf.getBegin(), 1);
+    assertEquals(sf.getEnd(), 437);
+    assertEquals(sf.getDescription(),
+            "Exon 2 for protein EMBLCDS:AAA91567.1");
+    assertEquals(sf.getFeatureGroup(), "EmblTest");
+    assertEquals(sf.getEnaLocation(), "join(7022..7502,1..437)");
+    assertEquals(sf.getPhase(), "0");
+    assertEquals(sf.getStrand(), 1);
+    assertEquals(sf.getValue("note"), "pGP7-D");
+    // second exon of circular DNA!
+    assertEquals(sf.getValue("exon number"), 2);
+    assertEquals(sf.getValue("product"), "hypothetical protein");
+    assertEquals(sf.getValue("transl_table"), "11");
+    
+    sf = features.get(1);
+    assertEquals(sf.getBegin(), 488);
+    assertEquals(sf.getEnd(), 1480);
+    assertEquals(sf.getDescription(),
+            "Exon 1 for protein EMBLCDS:AAA91568.1");
+    assertEquals(sf.getFeatureGroup(), "EmblTest");
+    assertEquals(sf.getEnaLocation(), "complement(488..1480)");
+    assertEquals(sf.getPhase(), "0");
+    assertEquals(sf.getStrand(), -1); // reverse strand!
+    assertEquals(sf.getValue("note"), "pGP8-D");
+    assertEquals(sf.getValue("exon number"), 1);
+    assertEquals(sf.getValue("product"), "hypothetical protein");
+    
+    sf = features.get(7);
+    assertEquals(sf.getBegin(), 6045);
+    assertEquals(sf.getEnd(), 6788);
+    assertEquals(sf.getDescription(),
+            "Exon 1 for protein EMBLCDS:AAA91574.1");
+    assertEquals(sf.getFeatureGroup(), "EmblTest");
+    assertEquals(sf.getEnaLocation(), "6045..6788");
+    assertEquals(sf.getPhase(), "0");
+    assertEquals(sf.getStrand(), 1);
+    assertEquals(sf.getValue("note"), "pGP6-D (gtg start codon)");
+    assertEquals(sf.getValue("exon number"), 1);
+    assertEquals(sf.getValue("product"), "hypothetical protein");
+    
+    /*
+     * CDS at 7022-7502 is the first exon of the circular DNA CDS
+     */
+    sf = features.get(8);
+    assertEquals(sf.getBegin(), 7022);
+    assertEquals(sf.getEnd(), 7502);
+    assertEquals(sf.getDescription(),
+            "Exon 1 for protein EMBLCDS:AAA91567.1");
+    assertEquals(sf.getFeatureGroup(), "EmblTest");
+    assertEquals(sf.getEnaLocation(), "join(7022..7502,1..437)");
+    assertEquals(sf.getPhase(), "0");
+    assertEquals(sf.getStrand(), 1);
+    assertEquals(sf.getValue("note"), "pGP7-D");
+    assertEquals(sf.getValue("exon number"), 1);
+    assertEquals(sf.getValue("product"), "hypothetical protein");
+
+    /*
+     * there are 4 'direct' (DR) dbrefs, and numerous CDS /db_xref entries,
+     * some of them (e.g. INTERPRO) duplicates; sample a few here
+     * Note DBRefEntry constructor capitalises source
+     */
+    List<DBRefEntry> dbrefs = seq.getDBRefs();
+    assertEquals(dbrefs.size(), 31);
+    // 1st DR line; note trailing period is removed
+    assertTrue(dbrefs.contains(new DBRefEntry("MD5", "0",
+            "d4c4942a634e3df4995fd5ac75c26a61")));
+    // the 4th DR line:
+    assertTrue(
+            dbrefs.contains(new DBRefEntry("EuropePMC", "0", "PMC87941")));
+    // from the first CDS feature; note canonicalisation to "UNIPROT"
+    assertTrue(dbrefs.contains(new DBRefEntry("GOA", "0", "P0CE19")));
+    assertTrue(dbrefs.contains(new DBRefEntry("UNIPROT", "0", "P0CE19")));
+    // from the last CDS feature
+    assertTrue(dbrefs.contains(new DBRefEntry("INTERPRO", "0", "IPR005350")));
+
+    // todo: mappings to, and sequences for, UNIPROT proteins
+  }
+}