1 package jalview.ext.ensembl;
3 import jalview.analysis.AlignmentUtils;
4 import jalview.datamodel.Alignment;
5 import jalview.datamodel.AlignmentI;
6 import jalview.datamodel.DBRefEntry;
7 import jalview.datamodel.DBRefSource;
8 import jalview.datamodel.Mapping;
9 import jalview.datamodel.SequenceFeature;
10 import jalview.datamodel.SequenceI;
11 import jalview.exceptions.JalviewException;
12 import jalview.io.FastaFile;
13 import jalview.io.FileParse;
14 import jalview.io.gff.SequenceOntology;
15 import jalview.schemes.ResidueProperties;
16 import jalview.util.DBRefUtils;
17 import jalview.util.MapList;
18 import jalview.util.MappingUtils;
19 import jalview.util.StringUtils;
21 import java.io.IOException;
22 import java.net.MalformedURLException;
24 import java.util.ArrayList;
25 import java.util.Arrays;
26 import java.util.Collections;
27 import java.util.Comparator;
28 import java.util.LinkedHashMap;
29 import java.util.List;
30 import java.util.Map.Entry;
33 * Base class for Ensembl sequence fetchers
37 public abstract class EnsemblSeqProxy extends EnsemblRestClient
39 protected static final String CONSEQUENCE_TYPE = "consequence_type";
41 protected static final String PARENT = "Parent";
43 protected static final String ID = "ID";
46 * this needs special handling, as it isA sequence_variant in the
47 * Sequence Ontology, but behaves in Ensembl as if it isA transcript
49 protected static final String NMD_VARIANT = "NMD_transcript_variant";
51 protected static final String NAME = "Name";
53 public enum EnsemblSeqType
56 * type=genomic for the full dna including introns
61 * type=cdna for transcribed dna including UTRs
66 * type=cds for coding dna excluding UTRs
71 * type=protein for the peptide product sequence
76 * the value of the 'type' parameter to fetch this version of
81 EnsemblSeqType(String t)
86 public String getType()
94 * A comparator to sort ranges into ascending start position order
96 private class RangeSorter implements Comparator<int[]>
100 RangeSorter(boolean forward)
106 public int compare(int[] o1, int[] o2)
108 return (forwards ? 1 : -1) * Integer.compare(o1[0], o2[0]);
116 public EnsemblSeqProxy()
121 * Makes the sequence queries to Ensembl's REST service and returns an
122 * alignment consisting of the returned sequences.
125 public AlignmentI getSequenceRecords(String query) throws Exception
127 long now = System.currentTimeMillis();
128 // TODO use a String... query vararg instead?
130 // danger: accession separator used as a regex here, a string elsewhere
131 // in this case it is ok (it is just a space), but (e.g.) '\' would not be
132 List<String> allIds = Arrays.asList(query
133 .split(getAccessionSeparator()));
134 AlignmentI alignment = null;
138 * execute queries, if necessary in batches of the
139 * maximum allowed number of ids
141 int maxQueryCount = getMaximumQueryCount();
142 for (int v = 0, vSize = allIds.size(); v < vSize; v += maxQueryCount)
144 int p = Math.min(vSize, v + maxQueryCount);
145 List<String> ids = allIds.subList(v, p);
148 alignment = fetchSequences(ids, alignment);
149 } catch (Throwable r)
152 String msg = "Aborting ID retrieval after " + v
153 + " chunks. Unexpected problem (" + r.getLocalizedMessage()
155 System.err.println(msg);
156 if (alignment != null)
158 break; // return what we got
162 throw new JalviewException(msg, r);
168 * fetch and transfer genomic sequence features,
169 * fetch protein product and add as cross-reference
171 for (String accId : allIds)
173 addFeaturesAndProduct(accId, alignment);
177 System.out.println(getClass().getName() + " took "
178 + (System.currentTimeMillis() - now) + "ms to fetch");
183 * Fetches Ensembl features using the /overlap REST endpoint, and adds them to
184 * the sequence in the alignment. Also fetches the protein product, maps it
185 * from the CDS features of the sequence, and saves it as a cross-reference of
191 protected void addFeaturesAndProduct(String accId, AlignmentI alignment)
193 if (alignment == null)
201 * get 'dummy' genomic sequence with exon, cds and variation features
203 SequenceI genomicSequence = null;
204 EnsemblOverlap gffFetcher = new EnsemblOverlap();
205 EnsemblFeatureType[] features = getFeaturesToFetch();
206 AlignmentI geneFeatures = gffFetcher.getSequenceRecords(accId,
208 if (geneFeatures.getHeight() > 0)
210 genomicSequence = geneFeatures.getSequenceAt(0);
212 if (genomicSequence != null)
215 * transfer features to the query sequence
217 SequenceI querySeq = alignment.findName(accId);
218 if (transferFeatures(accId, genomicSequence, querySeq))
222 * fetch and map protein product, and add it as a cross-reference
223 * of the retrieved sequence
225 addProteinProduct(querySeq);
228 } catch (IOException e)
230 System.err.println("Error transferring Ensembl features: "
236 * Returns those sequence feature types to fetch from Ensembl. We may want
237 * features either because they are of interest to the user, or as means to
238 * identify the locations of the sequence on the genomic sequence (CDS
239 * features identify CDS, exon features identify cDNA etc).
243 protected abstract EnsemblFeatureType[] getFeaturesToFetch();
246 * Fetches and maps the protein product, and adds it as a cross-reference of
247 * the retrieved sequence
249 protected void addProteinProduct(SequenceI querySeq)
251 String accId = querySeq.getName();
254 AlignmentI protein = new EnsemblProtein().getSequenceRecords(accId);
255 if (protein == null || protein.getHeight() == 0)
257 System.out.println("Failed to retrieve protein for " + accId);
260 SequenceI proteinSeq = protein.getSequenceAt(0);
263 * need dataset sequences (to be the subject of mappings)
265 proteinSeq.createDatasetSequence();
266 querySeq.createDatasetSequence();
268 MapList mapList = mapCdsToProtein(querySeq, proteinSeq);
271 Mapping map = new Mapping(proteinSeq.getDatasetSequence(), mapList);
272 DBRefEntry dbr = new DBRefEntry(getDbSource(), getDbVersion(),
274 querySeq.getDatasetSequence().addDBRef(dbr);
277 * compute peptide variants from dna variants and add as
278 * sequence features on the protein sequence ta-da
280 computeProteinFeatures(querySeq, proteinSeq, mapList);
282 } catch (Exception e)
285 .println(String.format("Error retrieving protein for %s: %s",
286 accId, e.getMessage()));
291 * Returns a mapping from dna to protein by inspecting sequence features of
292 * type "CDS" on the dna.
298 protected MapList mapCdsToProtein(SequenceI dnaSeq, SequenceI proteinSeq)
300 List<int[]> ranges = new ArrayList<int[]>(50);
302 int mappedDnaLength = getCdsRanges(dnaSeq, ranges);
304 int proteinLength = proteinSeq.getLength();
305 List<int[]> proteinRange = new ArrayList<int[]>();
306 int proteinStart = 1;
309 * incomplete start codon may mean X at start of peptide
310 * we ignore both for mapping purposes
312 if (proteinSeq.getCharAt(0) == 'X')
317 proteinRange.add(new int[] { proteinStart, proteinLength });
320 * dna length should map to protein (or protein plus stop codon)
322 int codesForResidues = mappedDnaLength / 3;
323 if (codesForResidues == proteinLength
324 || codesForResidues == (proteinLength + 1))
326 return new MapList(ranges, proteinRange, 3, 1);
332 * Adds CDS ranges to the ranges list, and returns the total length mapped.
334 * No need to worry about reverse strand dna here since the retrieved sequence
335 * is as transcribed (reverse complement for reverse strand), i.e in the same
336 * sense as the peptide.
342 protected int getCdsRanges(SequenceI dnaSeq, List<int[]> ranges)
344 SequenceFeature[] sfs = dnaSeq.getSequenceFeatures();
349 int mappedDnaLength = 0;
350 for (SequenceFeature sf : sfs)
353 * process a CDS feature (or a sub-type of CDS)
355 if (SequenceOntology.getInstance().isA(sf.getType(), SequenceOntology.CDS))
359 phase = Integer.parseInt(sf.getPhase());
360 } catch (NumberFormatException e)
365 * phase > 0 on first codon means 5' incomplete - skip to the start
366 * of the next codon; example ENST00000496384
368 int begin = sf.getBegin();
369 int end = sf.getEnd();
370 if (ranges.isEmpty() && phase > 0)
375 continue; // shouldn't happen?
378 ranges.add(new int[] { begin, end });
379 mappedDnaLength += Math.abs(end - begin) + 1;
382 return mappedDnaLength;
386 * Fetches sequences for the list of accession ids and adds them to the
387 * alignment. Returns the extended (or created) alignment.
392 * @throws JalviewException
393 * @throws IOException
395 protected AlignmentI fetchSequences(List<String> ids, AlignmentI alignment)
396 throws JalviewException, IOException
398 if (!isEnsemblAvailable())
401 throw new JalviewException("ENSEMBL Rest API not available.");
403 FileParse fp = getSequenceReader(ids);
404 FastaFile fr = new FastaFile(fp);
405 if (fr.hasWarningMessage())
407 System.out.println(String.format(
408 "Warning when retrieving %d ids %s\n%s", ids.size(),
409 ids.toString(), fr.getWarningMessage()));
411 else if (fr.getSeqs().size() != ids.size())
413 System.out.println(String.format(
414 "Only retrieved %d sequences for %d query strings", fr
415 .getSeqs().size(), ids.size()));
418 if (fr.getSeqs().size() == 1 && fr.getSeqs().get(0).getLength() == 0)
421 * POST request has returned an empty FASTA file e.g. for invalid id
423 throw new IOException("No data returned for " + ids);
426 if (fr.getSeqs().size() > 0)
428 AlignmentI seqal = new Alignment(
429 fr.getSeqsAsArray());
430 for (SequenceI sq:seqal.getSequences())
432 if (sq.getDescription() == null)
434 sq.setDescription(getDbName());
436 String name = sq.getName();
437 if (ids.contains(name)
438 || ids.contains(name.replace("ENSP", "ENST")))
440 DBRefUtils.parseToDbRef(sq, DBRefSource.ENSEMBL, "0", name);
443 if (alignment == null)
449 alignment.append(seqal);
456 * Returns the URL for the REST call
459 * @throws MalformedURLException
462 protected URL getUrl(List<String> ids) throws MalformedURLException
465 * a single id is included in the URL path
466 * multiple ids go in the POST body instead
468 StringBuffer urlstring = new StringBuffer(128);
469 urlstring.append(SEQUENCE_ID_URL);
472 urlstring.append("/").append(ids.get(0));
474 // @see https://github.com/Ensembl/ensembl-rest/wiki/Output-formats
475 urlstring.append("?type=").append(getSourceEnsemblType().getType());
476 urlstring.append(("&Accept=text/x-fasta"));
478 URL url = new URL(urlstring.toString());
483 * A sequence/id POST request currently allows up to 50 queries
485 * @see http://rest.ensembl.org/documentation/info/sequence_id_post
488 public int getMaximumQueryCount()
494 protected boolean useGetRequest()
500 protected String getRequestMimeType(boolean multipleIds)
502 return multipleIds ? "application/json" : "text/x-fasta";
506 protected String getResponseMimeType()
508 return "text/x-fasta";
513 * @return the configured sequence return type for this source
515 protected abstract EnsemblSeqType getSourceEnsemblType();
518 * Returns a list of [start, end] genomic ranges corresponding to the sequence
521 * The correspondence between the frames of reference is made by locating
522 * those features on the genomic sequence which identify the retrieved
523 * sequence. Specifically
525 * <li>genomic sequence is identified by "transcript" features with
526 * ID=transcript:transcriptId</li>
527 * <li>cdna sequence is identified by "exon" features with
528 * Parent=transcript:transcriptId</li>
529 * <li>cds sequence is identified by "CDS" features with
530 * Parent=transcript:transcriptId</li>
533 * The returned ranges are sorted to run forwards (for positive strand) or
534 * backwards (for negative strand). Aborts and returns null if both positive
535 * and negative strand are found (this should not normally happen).
537 * @param sourceSequence
540 * the start position of the sequence we are mapping to
543 protected MapList getGenomicRanges(SequenceI sourceSequence,
544 String accId, int start)
546 SequenceFeature[] sfs = sourceSequence.getSequenceFeatures();
553 * generously initial size for number of cds regions
554 * (worst case titin Q8WZ42 has c. 313 exons)
556 List<int[]> regions = new ArrayList<int[]>(100);
557 int mappedLength = 0;
558 int direction = 1; // forward
559 boolean directionSet = false;
561 for (SequenceFeature sf : sfs)
564 * accept the target feature type or a specialisation of it
565 * (e.g. coding_exon for exon)
567 if (identifiesSequence(sf, accId))
569 int strand = sf.getStrand();
571 if (directionSet && strand != direction)
573 // abort - mix of forward and backward
574 System.err.println("Error: forward and backward strand for "
582 * add to CDS ranges, semi-sorted forwards/backwards
586 regions.add(0, new int[] { sf.getEnd(), sf.getBegin() });
590 regions.add(new int[] { sf.getBegin(), sf.getEnd() });
592 mappedLength += Math.abs(sf.getEnd() - sf.getBegin() + 1);
597 * 'gene' sequence is contiguous so we can stop as soon as its
598 * identifying feature has been found
605 if (regions.isEmpty())
607 System.out.println("Failed to identify target sequence for " + accId
608 + " from genomic features");
613 * a final sort is needed since Ensembl returns CDS sorted within source
614 * (havana / ensembl_havana)
616 Collections.sort(regions, new RangeSorter(direction == 1));
618 List<int[]> to = new ArrayList<int[]>();
619 to.add(new int[] { start, start + mappedLength - 1 });
621 return new MapList(regions, to, 1, 1);
625 * Answers true if the sequence being retrieved may occupy discontiguous
626 * regions on the genomic sequence.
628 protected boolean isSpliceable()
634 * Returns true if the sequence feature marks positions of the genomic
635 * sequence feature which are within the sequence being retrieved. For
636 * example, an 'exon' feature whose parent is the target transcript marks the
637 * cdna positions of the transcript.
643 protected abstract boolean identifiesSequence(SequenceFeature sf,
647 * Transfers the sequence feature to the target sequence, locating its start
648 * and end range based on the mapping. Features which do not overlap the
649 * target sequence are ignored.
652 * @param targetSequence
654 * mapping from the sequence feature's coordinates to the target
657 protected void transferFeature(SequenceFeature sf,
658 SequenceI targetSequence, MapList mapping)
660 int start = sf.getBegin();
661 int end = sf.getEnd();
662 int[] mappedRange = mapping.locateInTo(start, end);
664 if (mappedRange != null)
666 SequenceFeature copy = new SequenceFeature(sf);
667 copy.setBegin(Math.min(mappedRange[0], mappedRange[1]));
668 copy.setEnd(Math.max(mappedRange[0], mappedRange[1]));
669 targetSequence.addSequenceFeature(copy);
672 * for sequence_variant, make an additional feature with consequence
674 if (SequenceOntology.getInstance().isSequenceVariant(sf.getType()))
676 String consequence = (String) sf.getValue(CONSEQUENCE_TYPE);
677 if (consequence != null)
679 SequenceFeature sf2 = new SequenceFeature("consequence",
680 consequence, copy.getBegin(), copy.getEnd(), 0f,
682 targetSequence.addSequenceFeature(sf2);
689 * Transfers features from sourceSequence to targetSequence
692 * @param sourceSequence
693 * @param targetSequence
694 * @return true if any features were transferred, else false
696 protected boolean transferFeatures(String accessionId,
697 SequenceI sourceSequence, SequenceI targetSequence)
699 if (sourceSequence == null || targetSequence == null)
704 SequenceFeature[] sfs = sourceSequence.getSequenceFeatures();
705 MapList mapping = getGenomicRanges(sourceSequence, accessionId,
706 targetSequence.getStart());
712 return transferFeatures(sfs, targetSequence, mapping, accessionId);
716 * Transfer features to the target sequence. The start/end positions are
717 * converted using the mapping. Features which do not overlap are ignored.
718 * Features whose parent is not the specified identifier are also ignored.
721 * @param targetSequence
726 protected boolean transferFeatures(SequenceFeature[] features,
727 SequenceI targetSequence, MapList mapping, String parentId)
729 final boolean forwardStrand = mapping.isFromForwardStrand();
732 * sort features by start position (descending if reverse strand)
733 * before transferring (in forwards order) to the target sequence
735 Arrays.sort(features, new Comparator<SequenceFeature>()
738 public int compare(SequenceFeature o1, SequenceFeature o2)
740 int c = Integer.compare(o1.getBegin(), o2.getBegin());
741 return forwardStrand ? c : -c;
745 boolean transferred = false;
746 for (SequenceFeature sf : features)
748 if (retainFeature(sf, parentId))
750 transferFeature(sf, targetSequence, mapping);
758 * Answers true if the feature type is one we want to keep for the sequence.
759 * Some features are only retrieved in order to identify the sequence range,
760 * and may then be discarded as redundant information (e.g. "CDS" feature for
763 @SuppressWarnings("unused")
764 protected boolean retainFeature(SequenceFeature sf, String accessionId)
766 return true; // override as required
770 * Answers true if the feature has a Parent which refers to the given
771 * accession id, or if the feature has no parent. Answers false if the
772 * feature's Parent is for a different accession id.
778 protected boolean featureMayBelong(SequenceFeature sf, String identifier)
780 String parent = (String) sf.getValue(PARENT);
781 // using contains to allow for prefix "gene:", "transcript:" etc
782 if (parent != null && !parent.contains(identifier))
784 // this genomic feature belongs to a different transcript
791 public String getDescription()
793 return "Ensembl " + getSourceEnsemblType().getType()
794 + " sequence with variant features";
798 * Returns a (possibly empty) list of features on the sequence which have the
799 * specified sequence ontology type (or a sub-type of it), and the given
800 * identifier as parent
807 protected List<SequenceFeature> findFeatures(SequenceI sequence,
808 String type, String parentId)
810 List<SequenceFeature> result = new ArrayList<SequenceFeature>();
812 SequenceFeature[] sfs = sequence.getSequenceFeatures();
814 SequenceOntology so = SequenceOntology.getInstance();
815 for (SequenceFeature sf :sfs) {
816 if (so.isA(sf.getType(), type))
818 String parent = (String) sf.getValue(PARENT);
819 if (parent.equals(parentId))
830 * Maps exon features from dna to protein, and computes variants in peptide
831 * product generated by variants in dna, and adds them as sequence_variant
832 * features on the protein sequence. Returns the number of variant features
837 * @param dnaToProtein
839 static int computeProteinFeatures(SequenceI dnaSeq,
840 SequenceI peptide, MapList dnaToProtein)
842 while (dnaSeq.getDatasetSequence() != null)
844 dnaSeq = dnaSeq.getDatasetSequence();
846 while (peptide.getDatasetSequence() != null)
848 peptide = peptide.getDatasetSequence();
851 AlignmentUtils.transferFeatures(dnaSeq, peptide, dnaToProtein,
852 SequenceOntology.EXON);
854 LinkedHashMap<Integer, String[][]> variants = buildDnaVariantsMap(
855 dnaSeq, dnaToProtein);
858 * scan codon variations, compute peptide variants and add to peptide sequence
861 for (Entry<Integer, String[][]> variant : variants.entrySet())
863 int peptidePos = variant.getKey();
864 String[][] codonVariants = variant.getValue();
865 String residue = String.valueOf(peptide.getCharAt(peptidePos - 1)); // 0-based
866 List<String> peptideVariants = computePeptideVariants(codonVariants,
868 if (!peptideVariants.isEmpty())
870 String desc = StringUtils.listToDelimitedString(peptideVariants,
872 SequenceFeature sf = new SequenceFeature(
873 SequenceOntology.SEQUENCE_VARIANT, desc, peptidePos,
874 peptidePos, 0f, null);
875 peptide.addSequenceFeature(sf);
883 * Builds a map whose key is position in the protein sequence, and value is an
884 * array of all variants for the coding codon positions
887 * @param dnaToProtein
890 static LinkedHashMap<Integer, String[][]> buildDnaVariantsMap(
891 SequenceI dnaSeq, MapList dnaToProtein)
894 * map from peptide position to all variant features of the codon for it
895 * LinkedHashMap ensures we add the peptide features in sequence order
897 LinkedHashMap<Integer, String[][]> variants = new LinkedHashMap<Integer, String[][]>();
898 SequenceOntology so = SequenceOntology.getInstance();
900 SequenceFeature[] dnaFeatures = dnaSeq.getSequenceFeatures();
901 if (dnaFeatures == null)
906 int dnaStart = dnaSeq.getStart();
907 int[] lastCodon = null;
908 int lastPeptidePostion = 0;
911 * build a map of codon variations for peptides
913 for (SequenceFeature sf : dnaFeatures)
915 int dnaCol = sf.getBegin();
916 if (dnaCol != sf.getEnd())
918 // not handling multi-locus variant features
921 if (so.isSequenceVariant(sf.getType()))
923 int[] mapsTo = dnaToProtein.locateInTo(dnaCol, dnaCol);
926 // feature doesn't lie within coding region
929 int peptidePosition = mapsTo[0];
930 String[][] codonVariants = variants.get(peptidePosition);
931 if (codonVariants == null)
933 codonVariants = new String[3][];
934 variants.put(peptidePosition, codonVariants);
938 * extract dna variants to a string array
940 String alls = (String) sf.getValue("alleles");
945 String[] alleles = alls.split(",");
948 * get this peptides codon positions e.g. [3, 4, 5] or [4, 7, 10]
950 int[] codon = peptidePosition == lastPeptidePostion ? lastCodon
951 : MappingUtils.flattenRanges(dnaToProtein.locateInFrom(
952 peptidePosition, peptidePosition));
953 lastPeptidePostion = peptidePosition;
957 * save nucleotide (and this variant) for each codon position
959 for (int codonPos = 0; codonPos < 3; codonPos++)
961 String nucleotide = String.valueOf(dnaSeq
962 .getCharAt(codon[codonPos] - dnaStart));
963 if (codon[codonPos] == dnaCol)
966 * record current dna base and its alleles
968 String[] dnaVariants = new String[alleles.length + 1];
969 dnaVariants[0] = nucleotide;
970 System.arraycopy(alleles, 0, dnaVariants, 1, alleles.length);
971 codonVariants[codonPos] = dnaVariants;
973 else if (codonVariants[codonPos] == null)
976 * record current dna base only
977 * (at least until we find any variation and overwrite it)
979 codonVariants[codonPos] = new String[] { nucleotide };
988 * Returns a sorted, non-redundant list of all peptide translations generated
989 * by the given dna variants, excluding the current residue value
991 * @param codonVariants
992 * an array of base values (acgtACGT) for codon positions 1, 2, 3
994 * the current residue translation
997 static List<String> computePeptideVariants(
998 String[][] codonVariants, String residue)
1000 List<String> result = new ArrayList<String>();
1001 for (String base1 : codonVariants[0])
1003 for (String base2 : codonVariants[1])
1005 for (String base3 : codonVariants[2])
1007 String codon = base1 + base2 + base3;
1008 // TODO: report frameshift/insertion/deletion
1009 // and multiple-base variants?!
1010 String peptide = codon.contains("-") ? "-" : ResidueProperties
1011 .codonTranslate(codon);
1012 if (peptide != null && !result.contains(peptide)
1013 && !peptide.equalsIgnoreCase(residue))
1015 result.add(peptide);
1022 * sort alphabetically with STOP at the end
1024 Collections.sort(result, new Comparator<String>()
1028 public int compare(String o1, String o2)
1030 if ("STOP".equals(o1))
1034 else if ("STOP".equals(o2))
1040 return o1.compareTo(o2);
1048 * Answers true if the feature type is either 'NMD_transcript_variant' or
1049 * 'transcript' or one of its sub-types in the Sequence Ontology. This is
1050 * needed because NMD_transcript_variant behaves like 'transcript' in Ensembl
1051 * although strictly speaking it is not (it is a sub-type of
1052 * sequence_variant).
1054 * @param featureType
1057 public static boolean isTranscript(String featureType)
1059 return NMD_VARIANT.equals(featureType)
1060 || SequenceOntology.getInstance().isA(featureType, SequenceOntology.TRANSCRIPT);