Trm 30823 UniParc proteome redundant fasta

core-domain/src/main/java/org/uniprot/core/uniparc/UniParcCrossReference.java

@@ -11,6 +11,8 @@
  */
 public interface UniParcCrossReference extends CrossReference<UniParcDatabase> {
 
+    public static final String PROPERTY_SOURCES = "sources";
+
     int getVersionI();
 
     Integer getVersion();

core-domain/src/main/java/org/uniprot/core/uniparc/UniParcDatabase.java

@@ -6,92 +6,99 @@
 import org.uniprot.core.util.EnumDisplay;
 
 public enum UniParcDatabase implements Database, EnumDisplay {
-    EG_BACTERIA(900, "EnsemblBacteria", true, "https://www.ensemblgenomes.org/id/%id"),
-    EG_FUNGI(1000, "EnsemblFungi", true, "https://www.ensemblgenomes.org/id/%id"),
-    EG_METAZOA(1100, "EnsemblMetazoa", true, "https://www.ensemblgenomes.org/id/%id"),
-    EG_PLANTS(1200, "EnsemblPlants", true, "https://www.ensemblgenomes.org/id/%id"),
-    EG_PROTISTS(1300, "EnsemblProtists", true, "https://www.ensemblgenomes.org/id/%id"),
-
-    EMBL(300, "EMBL", true, "https://www.ebi.ac.uk/ena/browser/view/%id"),
-    EMBL_CON(400, "EMBL_CON", true, "https://www.ebi.ac.uk/ena/browser/view/%id"),
-    EMBL_TPA(500, "EMBL_TPA", false, "https://www.ebi.ac.uk/ena/browser/view/%id"),
-    EMBL_TSA(600, "EMBL_TSA", true, "https://www.ebi.ac.uk/ena/browser/view/%id"),
-    EMBLWGS(700, "EMBLWGS", true, "https://www.ebi.ac.uk/ena/browser/view/%id"),
-
-    ENSEMBL_VERTEBRATE(800, "Ensembl", true, "https://www.ensembl.org/id/%id"),
-    ENSEMBL_RAPID(1350, "EnsemblRapid", true, "https://rapid.ensembl.org/id/%id"),
-
-    EPO(1400, "EPO", true, "https://www.ebi.ac.uk/Tools/dbfetch/dbfetch?db=epo_prt&id=%id"),
-    FLYBASE(1500, "FlyBase", true, "https://flybase.org/reports/%id.html"),
+    EG_BACTERIA(900, "EnsemblBacteria", true, true, "https://www.ensemblgenomes.org/id/%id"),
+    EG_FUNGI(1000, "EnsemblFungi", true, true, "https://www.ensemblgenomes.org/id/%id"),
+    EG_METAZOA(1100, "EnsemblMetazoa", true, true, "https://www.ensemblgenomes.org/id/%id"),
+    EG_PLANTS(1200, "EnsemblPlants", true, true, "https://www.ensemblgenomes.org/id/%id"),
+    EG_PROTISTS(1300, "EnsemblProtists", true, true, "https://www.ensemblgenomes.org/id/%id"),
+
+    EMBL(300, "EMBL", true, true, "https://www.ebi.ac.uk/ena/browser/view/%id"),
+    EMBL_CON(400, "EMBL_CON", true, true, "https://www.ebi.ac.uk/ena/browser/view/%id"),
+    EMBL_TPA(500, "EMBL_TPA", false, true, "https://www.ebi.ac.uk/ena/browser/view/%id"),
+    EMBL_TSA(600, "EMBL_TSA", true, true, "https://www.ebi.ac.uk/ena/browser/view/%id"),
+    EMBLWGS(700, "EMBLWGS", true, true, "https://www.ebi.ac.uk/ena/browser/view/%id"),
+
+    ENSEMBL_VERTEBRATE(800, "Ensembl", true, true, "https://www.ensembl.org/id/%id"),
+    ENSEMBL_RAPID(1350, "EnsemblRapid", true, true, "https://rapid.ensembl.org/id/%id"),
+
+    EPO(1400, "EPO", true, false, "https://www.ebi.ac.uk/Tools/dbfetch/dbfetch?db=epo_prt&id=%id"),
+    FLYBASE(1500, "FlyBase", true,false, "https://flybase.org/reports/%id.html"),
     FUSION_GDB(
             1550,
             "FusionGDB",
             true,
+            false,
             "https://compbio.uth.edu/FusionGDB2/gene_search_result.cgi?type=quick_search&quick_search=%id"),
 
-    H_INV(1600, "H-InvDB", false),
-    IPI(1700, "IPI", false),
+    H_INV(1600, "H-InvDB",false, false),
+    IPI(1700, "IPI",false, false),
 
-    JPO(1800, "JPO", true, "https://www.ebi.ac.uk/Tools/dbfetch/dbfetch?db=jpo_prt&id=%id"),
-    KIPO(1900, "KIPO", true, "https://www.ebi.ac.uk/Tools/dbfetch/dbfetch?db=kipo_prt&id=%id"),
-    PATRIC(2000, "PATRIC", true, "https://www.patricbrc.org/view/Feature/%id"),
+    JPO(1800, "JPO", true,false, "https://www.ebi.ac.uk/Tools/dbfetch/dbfetch?db=jpo_prt&id=%id"),
+    KIPO(1900, "KIPO", true,false, "https://www.ebi.ac.uk/Tools/dbfetch/dbfetch?db=kipo_prt&id=%id"),
+    PATRIC(2000, "PATRIC", true,false, "https://www.patricbrc.org/view/Feature/%id"),
     PDB(
             2100,
             "PDB",
             true,
+            false,
             "https://www.ebi.ac.uk/pdbe/entry/pdb/%id"), // need to remove the chain, eg "4q8n_A",
     // just use "4q8n" as id
-    PIR(2200, "PIR", false),
+    PIR(2200, "PIR",false, false),
 
-    PIRARC(2300, "PIRARC", false),
-    PRF(2400, "PRF", false, "http://www.prf.or.jp/cgi-bin/seqget.pl?id=%id"),
-    REFSEQ(2500, "RefSeq", true, "https://www.ncbi.nlm.nih.gov/protein/%id"),
-    REMTREMBL(2600, "REMTREMBL", false),
+    PIRARC(2300, "PIRARC",false, false),
+    PRF(2400, "PRF",false, false, "http://www.prf.or.jp/cgi-bin/seqget.pl?id=%id"),
+    REFSEQ(2500, "RefSeq", true, true,"https://www.ncbi.nlm.nih.gov/protein/%id"),
+    REMTREMBL(2600, "REMTREMBL",false, false),
     SEED(
             2700,
             "SEED",
             true,
+            false,
             "https://pubseed.theseed.org/seedviewer.cgi?page=Annotation&feature=%id"),
 
-    SGD(2800, "SGD", true, "https://www.yeastgenome.org/locus/%id"),
-    SWISSPROT(100, "UniProtKB/Swiss-Prot", true, "https://www.uniprot.org/uniprot/%id"),
+    SGD(2800, "SGD", true, false,"https://www.yeastgenome.org/locus/%id"),
+    SWISSPROT(100, "UniProtKB/Swiss-Prot", true, false,"https://www.uniprot.org/uniprot/%id"),
     SWISSPROT_VARSPLIC(
             200,
             "UniProtKB/Swiss-Prot protein isoforms",
             true,
+            false,
             "https://www.uniprot.org/uniprot/%id"), // swissprot isoform
     TAIR_ARABIDOPSIS(
             2900,
             "TAIR",
             true,
+            false,
             "https://www.arabidopsis.org/servlets/TairObject?type=aa_sequence&name=%id"),
-    TREMBL(100, "UniProtKB/TrEMBL", true, "https://www.uniprot.org/uniprot/%id"),
+    TREMBL(100, "UniProtKB/TrEMBL", true, false, "https://www.uniprot.org/uniprot/%id"),
 
-    TREMBLNEW(3000, "TREMBLNEW", false),
-    TREMBL_VARSPLIC(3100, "TREMBL_VARSPLIC", false),
-    TROME(3200, "TROME", true), // no link
-    UNIMES(3300, "UNIMES", false),
-    USPTO(3400, "USPTO", true, "https://www.ebi.ac.uk/Tools/dbfetch/dbfetch?db=uspto_prt&id=%id"),
+    TREMBLNEW(3000, "TREMBLNEW", false, false),
+    TREMBL_VARSPLIC(3100, "TREMBL_VARSPLIC", false, false),
+    TROME(3200, "TROME", true, false), // no link
+    UNIMES(3300, "UNIMES", false, false),
+    USPTO(3400, "USPTO", true,  false, "https://www.ebi.ac.uk/Tools/dbfetch/dbfetch?db=uspto_prt&id=%id"),
 
-    VECTORBASE(3500, "VectorBase", false),
-    VEGA(3600, "VEGA", true, "https://vega.sanger.ac.uk/id/%id"),
-    WORMBASE_PARASITE(3700, "WBParaSite", true, "https://parasite.wormbase.org/id/%id"),
-    WORMBASE(3800, "WormBase", true, "https://wormbase.org/db/seq/protein?name=%id;class=CDS");
+    VECTORBASE(3500, "VectorBase", false, false),
+    VEGA(3600, "VEGA", true, false,  "https://vega.sanger.ac.uk/id/%id"),
+    WORMBASE_PARASITE(3700, "WBParaSite", true, true, "https://parasite.wormbase.org/id/%id"),
+    WORMBASE(3800, "WormBase", true, true,"https://wormbase.org/db/seq/protein?name=%id;class=CDS");
 
     private final String displayName;
     private final boolean alive;
     private final String url;
     private final int index;
+    private final boolean source;
 
-    UniParcDatabase(int index, String displayName, boolean alive) {
-        this(index, displayName, alive, "");
+    UniParcDatabase(int index, String displayName, boolean alive, boolean source) {
+        this(index, displayName, alive, source, "");
     }
 
-    UniParcDatabase(int index, String displayName, boolean alive, String url) {
+    UniParcDatabase(int index, String displayName, boolean alive, boolean source,String url) {
         this.index = index;
         this.displayName = displayName;
         this.alive = alive;
         this.url = url;
+        this.source = source;
     }
 
     public int getIndex() {
@@ -110,6 +117,10 @@ public String getUrl() {
         return url;
     }
 
+    public boolean isSource() {
+        return source;
+    }
+
     public static @Nonnull UniParcDatabase typeOf(@Nonnull String displayName) {
         return EnumDisplay.typeOf(displayName, UniParcDatabase.class);
     }

core-domain/src/test/java/org/uniprot/core/uniparc/UniParcDatabaseTest.java

@@ -64,6 +64,16 @@ void hasNoUrl() {
         assertEquals("", UniParcDatabase.VECTORBASE.getUrl());
     }
 
+    @Test
+    void isSource() {
+        assertTrue( UniParcDatabase.EMBL.isSource());
+    }
+
+    @Test
+    void isNotSource() {
+        assertFalse( UniParcDatabase.FLYBASE.isSource());
+    }
+
     @Test
     void canGetIndex() {
         assertEquals(100, UniParcDatabase.SWISSPROT.getIndex());

core-parser/src/main/java/org/uniprot/core/parser/fasta/FastaUtils.java

@@ -0,0 +1,17 @@
+package org.uniprot.core.parser.fasta;
+
+public class FastaUtils {
+
+    public static String parseSequence(String sequence) {
+        StringBuilder sb = new StringBuilder();
+        int columnCounter = 0;
+        for (char c : sequence.toCharArray()) {
+            if (columnCounter % 60 == 0 && columnCounter > 0) {
+                sb.append("\n");
+            }
+            sb.append(c);
+            columnCounter++;
+        }
+        return sb.toString();
+    }
+}

core-parser/src/main/java/org/uniprot/core/parser/fasta/UniRefFastaParser.java

@@ -4,6 +4,8 @@
 import org.uniprot.core.uniref.UniRefEntry;
 import org.uniprot.core.uniref.UniRefEntryLight;
 
+import static org.uniprot.core.parser.fasta.FastaUtils.parseSequence;
+
 /**
  * @author jluo
  * @date: 22 Aug 2019
@@ -15,30 +17,14 @@ private UniRefFastaParser() {}
     public static String toFasta(UniRefEntryLight entry) {
         StringBuilder sb = new StringBuilder();
         sb.append(getHeader(entry)).append("\n");
-        int columnCounter = 0;
-        String sequence = entry.getRepresentativeMember().getSequence().getValue();
-        for (char c : sequence.toCharArray()) {
-            if (columnCounter % 60 == 0 && columnCounter > 0) {
-                sb.append("\n");
-            }
-            sb.append(c);
-            columnCounter++;
-        }
+        sb.append(parseSequence(entry.getRepresentativeMember().getSequence().getValue()));
         return sb.toString();
     }
 
     public static String toFasta(UniRefEntry entry) {
         StringBuilder sb = new StringBuilder();
         sb.append(getHeader(entry)).append("\n");
-        String sequence = entry.getRepresentativeMember().getSequence().getValue();
-        int columnCounter = 0;
-        for (char c : sequence.toCharArray()) {
-            if (columnCounter % 60 == 0 && columnCounter > 0) {
-                sb.append("\n");
-            }
-            columnCounter++;
-            sb.append(c);
-        }
+        sb.append(parseSequence(entry.getRepresentativeMember().getSequence().getValue()));
         return sb.toString();
     }
 

core-parser/src/main/java/org/uniprot/core/parser/fasta/UniParcFastaParser.java → core-parser/src/main/java/org/uniprot/core/parser/fasta/uniparc/UniParcFastaParser.java

@@ -1,10 +1,11 @@
-package org.uniprot.core.parser.fasta;
+package org.uniprot.core.parser.fasta.uniparc;
 
 import org.uniprot.core.Sequence;
 import org.uniprot.core.uniparc.UniParcCrossReference;
 import org.uniprot.core.uniparc.UniParcEntry;
 import org.uniprot.core.uniparc.UniParcEntryLight;
 
+import static org.uniprot.core.parser.fasta.FastaUtils.*;
 import static org.uniprot.core.uniparc.impl.UniParcEntryLightBuilder.HAS_ACTIVE_CROSS_REF;
 
 /**
@@ -14,6 +15,7 @@
 public class UniParcFastaParser {
     private UniParcFastaParser(){}
 
+
     public static String toFasta(UniParcEntry entry) {
         String status = "active";
         boolean isActive = entry.getUniParcCrossReferences()
@@ -39,14 +41,7 @@ private static String getFastaString(String entry, String status, Sequence seque
         sb.append(">").append(entry).append(" ");
         sb.append("status=").append(status);
         sb.append("\n");
-        int columnCounter = 0;
-        for (char c : sequence.getValue().toCharArray()) {
-            if (columnCounter % 60 == 0 && columnCounter > 0) {
-                sb.append("\n");
-            }
-            sb.append(c);
-            columnCounter++;
-        }
+        sb.append(parseSequence(sequence.getValue()));
         return sb.toString();
     }
 }