1 package jalview.io.gff;
3 import jalview.datamodel.AlignedCodonFrame;
4 import jalview.datamodel.AlignmentI;
5 import jalview.datamodel.MappingType;
6 import jalview.datamodel.SequenceFeature;
7 import jalview.datamodel.SequenceI;
8 import jalview.util.MapList;
9 import jalview.util.StringUtils;
11 import java.io.IOException;
12 import java.util.List;
16 * Base class with generic / common functionality for processing GFF3 data.
17 * Override this as required for any specialisations resulting from
18 * peculiarities of GFF3 generated by particular tools.
20 public class Gff3Helper extends GffHelperBase
22 protected static final String TARGET = "Target";
24 protected static final String ID = "ID";
26 private static final String NAME = "Name";
29 * GFF3 uses '=' to delimit name/value pairs in column 9, and comma to
30 * separate multiple values for a name
35 public static Map<String, List<String>> parseNameValuePairs(String text)
37 return parseNameValuePairs(text, ";", '=', ",");
41 * Process one GFF feature line (as modelled by SequenceFeature)
44 * the sequence with which this feature is associated
46 * the sequence feature with ATTRIBUTES property containing any
47 * additional attributes
49 * the alignment we are adding GFF to
51 * any new sequences referenced by the GFF
52 * @param relaxedIdMatching
53 * if true, match word tokens in sequence names
54 * @return true if the sequence feature should be added to the sequence, else
55 * false (i.e. it has been processed in another way e.g. to generate a
60 public SequenceFeature processGff(SequenceI seq, String[] gff,
61 AlignmentI align, List<SequenceI> newseqs,
62 boolean relaxedIdMatching) throws IOException
64 SequenceFeature sf = null;
68 String soTerm = gff[TYPE_COL];
69 String atts = gff[ATTRIBUTES_COL];
70 Map<String, List<String>> attributes = parseNameValuePairs(atts);
72 SequenceOntologyI so = SequenceOntologyFactory.getInstance();
73 if (so.isA(soTerm, SequenceOntologyI.PROTEIN_MATCH))
75 sf = processProteinMatch(attributes, seq, gff, align, newseqs,
78 else if (so.isA(soTerm, SequenceOntologyI.NUCLEOTIDE_MATCH))
80 sf = processNucleotideMatch(attributes, seq, gff, align,
81 newseqs, relaxedIdMatching);
85 sf = buildSequenceFeature(gff, attributes);
91 * fall back on generating a sequence feature with no special processing
93 sf = buildSequenceFeature(gff, null);
100 * Processes one GFF3 nucleotide (e.g. cDNA to genome) match.
103 * parsed GFF column 9 key/value(s)
105 * the sequence the GFF feature is on
107 * the GFF column data
109 * the alignment the sequence belongs to, where any new mappings
112 * a list of new 'virtual sequences' generated while parsing GFF
113 * @param relaxedIdMatching
114 * if true allow fuzzy search for a matching target sequence
115 * @return a sequence feature, if one should be added to the sequence, else
117 * @throws IOException
119 protected SequenceFeature processNucleotideMatch(
120 Map<String, List<String>> attributes, SequenceI seq,
121 String[] gffColumns, AlignmentI align, List<SequenceI> newseqs,
122 boolean relaxedIdMatching)
125 String strand = gffColumns[STRAND_COL];
128 * (For now) we don't process mappings from reverse complement ; to do
129 * this would require (a) creating a virtual sequence placeholder for
130 * the reverse complement (b) resolving the sequence by its id from some
131 * source (GFF ##FASTA or other) (c) creating the reverse complement
132 * sequence (d) updating the mapping to be to the reverse complement
134 if ("-".equals(strand))
137 .println("Skipping mapping from reverse complement as not yet supported");
141 List<String> targets = attributes.get(TARGET);
144 System.err.println("'Target' missing in GFF");
149 * Typically we only expect one Target per GFF line, but this can handle
150 * multiple matches, to the same or different sequences (e.g. dna variants)
152 for (String target : targets)
155 * Process "seqid start end [strand]"
157 String[] tokens = target.split(" ");
158 if (tokens.length < 3)
160 System.err.println("Incomplete Target: " + target);
165 * Locate the mapped sequence in the alignment, or as a
166 * (new or existing) virtual sequence in the newseqs list
168 String targetId = findTargetId(tokens[0], attributes);
169 SequenceI mappedSequence1 = findSequence(targetId, align,
170 newseqs, relaxedIdMatching);
171 SequenceI mappedSequence = mappedSequence1;
172 if (mappedSequence == null)
178 * get any existing mapping for these sequences (or start one),
179 * and add this mapped range
181 AlignedCodonFrame acf = getMapping(align, seq, mappedSequence);
185 int toStart = Integer.parseInt(tokens[1]);
186 int toEnd = Integer.parseInt(tokens[2]);
187 if (tokens.length > 3 && "-".equals(tokens[3]))
189 // mapping to reverse strand - swap start/end
195 int fromStart = Integer.parseInt(gffColumns[START_COL]);
196 int fromEnd = Integer.parseInt(gffColumns[END_COL]);
197 MapList mapping = constructMappingFromAlign(fromStart, fromEnd,
199 MappingType.NucleotideToNucleotide);
203 acf.addMap(seq, mappedSequence, mapping);
204 align.addCodonFrame(acf);
206 } catch (NumberFormatException nfe)
208 System.err.println("Invalid start or end in Target " + target);
212 SequenceFeature sf = buildSequenceFeature(gffColumns, attributes);
217 * Returns the target sequence id extracted from the GFF name/value pairs.
218 * Default (standard behaviour) is the first token for "Target". This may be
219 * overridden where tools report this in a non-standard way.
222 * first token of a "Target" value from GFF column 9, typically
225 * a map with all parsed column 9 attributes
228 @SuppressWarnings("unused")
229 protected String findTargetId(String target, Map<String, List<String>> set)
235 * Processes one GFF 'protein_match'; fields of interest are
237 * <li>feature group - the database reporting a match e.g. Pfam</li>
238 * <li>Name - the matched entry's accession id in the database</li>
239 * <li>ID - a sequence identifier for the matched region (which may be
240 * appended as FASTA in the GFF file)</li>
244 * parsed GFF column 9 key/value(s)
246 * the sequence the GFF feature is on
248 * the sequence feature holding GFF data
250 * the alignment the sequence belongs to, where any new mappings
253 * a list of new 'virtual sequences' generated while parsing GFF
254 * @param relaxedIdMatching
255 * if true allow fuzzy search for a matching target sequence
256 * @return the (real or virtual) sequence(s) mapped to by this match
257 * @throws IOException
259 protected SequenceFeature processProteinMatch(
260 Map<String, List<String>> set, SequenceI seq,
261 String[] gffColumns, AlignmentI align, List<SequenceI> newseqs,
262 boolean relaxedIdMatching)
264 // This is currently tailored to InterProScan GFF output:
265 // ID holds the ID of the matched sequence, Target references the
266 // query sequence; this looks wrong, as ID should just be the GFF internal
267 // ID of the GFF feature, while Target would normally reference the matched
269 // TODO refactor as needed if other protein-protein GFF varies
271 SequenceFeature sf = buildSequenceFeature(gffColumns, set);
274 * locate the mapped sequence in the alignment, or as a
275 * (new or existing) virtual sequence in the newseqs list
277 List<String> targets = set.get(TARGET);
280 for (String target : targets)
283 SequenceI mappedSequence1 = findSequence(findTargetId(target, set), align,
284 newseqs, relaxedIdMatching);
285 SequenceI mappedSequence = mappedSequence1;
286 if (mappedSequence == null)
292 * give the mapped sequence a copy of the sequence feature, with
293 * start/end range adjusted
295 SequenceFeature sf2 = new SequenceFeature(sf);
297 int sequenceFeatureLength = 1 + sf.getEnd() - sf.getBegin();
298 sf2.setEnd(sequenceFeatureLength);
299 mappedSequence.addSequenceFeature(sf2);
302 * add a property to the mapped sequence so that it can eventually be
303 * renamed with its qualified accession id; renaming has to wait until
304 * all sequence reference resolution is complete
306 String accessionId = StringUtils.listToDelimitedString(
308 if (accessionId.length() > 0)
310 String database = sf.getType(); // TODO InterProScan only??
311 String qualifiedAccId = database + "|" + accessionId;
312 sf2.setValue(RENAME_TOKEN, qualifiedAccId);
316 * get any existing mapping for these sequences (or start one),
317 * and add this mapped range
319 AlignedCodonFrame alco = getMapping(align, seq, mappedSequence);
320 int[] from = new int[] { sf.getBegin(), sf.getEnd() };
321 int[] to = new int[] { 1, sequenceFeatureLength };
322 MapList mapping = new MapList(from, to, 1, 1);
324 alco.addMap(seq, mappedSequence, mapping);
325 align.addCodonFrame(alco);
333 * Return '=' as the name-value separator used in column 9 attributes.
336 protected char getNameValueSeparator()
342 * Modifies the default SequenceFeature in order to set the Target sequence id
346 protected SequenceFeature buildSequenceFeature(String[] gff,
347 Map<String, List<String>> attributes)
349 SequenceFeature sf = super.buildSequenceFeature(gff, attributes);
350 String desc = getDescription(sf, attributes);
353 sf.setDescription(desc);
359 * Apply heuristic rules to try to get the most useful feature description
365 protected String getDescription(SequenceFeature sf,
366 Map<String, List<String>> attributes)
369 String target = (String) sf.getValue(TARGET);
372 desc = target.split(" ")[0];
375 SequenceOntologyI so = SequenceOntologyFactory.getInstance();
376 String type = sf.getType();
377 if (so.isA(type, SequenceOntologyI.SEQUENCE_VARIANT))
380 * Ensembl returns dna variants as 'alleles'
382 desc = StringUtils.listToDelimitedString(
383 attributes.get("alleles"), ",");
387 * extract 'Name' for a transcript (to show gene name)
388 * or an exon (so 'colour by label' shows exon boundaries)
390 if (SequenceOntologyI.NMD_TRANSCRIPT_VARIANT.equals(type)
391 || so.isA(type, SequenceOntologyI.TRANSCRIPT)
392 || so.isA(type, SequenceOntologyI.EXON))
394 desc = StringUtils.listToDelimitedString(attributes.get("Name"), ",");
398 * if the above fails, try ID
402 desc = (String) sf.getValue(ID);