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 if (SequenceOntology.getInstance().isProteinMatch(soTerm))
74 sf = processProteinMatch(attributes, seq, gff, align,
75 newseqs, relaxedIdMatching);
77 else if (SequenceOntology.getInstance().isNucleotideMatch(soTerm))
79 sf = processNucleotideMatch(attributes, seq, gff, align,
80 newseqs, relaxedIdMatching);
84 sf = buildSequenceFeature(gff, attributes);
90 * fall back on generating a sequence feature with no special processing
92 sf = buildSequenceFeature(gff, null);
99 * Processes one GFF3 nucleotide (e.g. cDNA to genome) match.
102 * parsed GFF column 9 key/value(s)
104 * the sequence the GFF feature is on
106 * the GFF column data
108 * the alignment the sequence belongs to, where any new mappings
111 * a list of new 'virtual sequences' generated while parsing GFF
112 * @param relaxedIdMatching
113 * if true allow fuzzy search for a matching target sequence
114 * @return a sequence feature, if one should be added to the sequence, else
116 * @throws IOException
118 protected SequenceFeature processNucleotideMatch(
119 Map<String, List<String>> attributes, SequenceI seq,
120 String[] gffColumns, AlignmentI align, List<SequenceI> newseqs,
121 boolean relaxedIdMatching)
124 String strand = gffColumns[STRAND_COL];
127 * (For now) we don't process mappings from reverse complement ; to do
128 * this would require (a) creating a virtual sequence placeholder for
129 * the reverse complement (b) resolving the sequence by its id from some
130 * source (GFF ##FASTA or other) (c) creating the reverse complement
131 * sequence (d) updating the mapping to be to the reverse complement
133 if ("-".equals(strand))
136 .println("Skipping mapping from reverse complement as not yet supported");
140 List<String> targets = attributes.get(TARGET);
143 System.err.println("'Target' missing in GFF");
148 * Typically we only expect one Target per GFF line, but this can handle
149 * multiple matches, to the same or different sequences (e.g. dna variants)
151 for (String target : targets)
154 * Process "seqid start end [strand]"
156 String[] tokens = target.split(" ");
157 if (tokens.length < 3)
159 System.err.println("Incomplete Target: " + target);
164 * Locate the mapped sequence in the alignment, or as a
165 * (new or existing) virtual sequence in the newseqs list
167 String targetId = findTargetId(tokens[0], attributes);
168 SequenceI mappedSequence1 = findSequence(targetId, align,
169 newseqs, relaxedIdMatching);
170 SequenceI mappedSequence = mappedSequence1;
171 if (mappedSequence == null)
177 * get any existing mapping for these sequences (or start one),
178 * and add this mapped range
180 AlignedCodonFrame acf = getMapping(align, seq, mappedSequence);
184 int toStart = Integer.parseInt(tokens[1]);
185 int toEnd = Integer.parseInt(tokens[2]);
186 if (tokens.length > 3 && "-".equals(tokens[3]))
188 // mapping to reverse strand - swap start/end
194 int fromStart = Integer.parseInt(gffColumns[START_COL]);
195 int fromEnd = Integer.parseInt(gffColumns[END_COL]);
196 MapList mapping = constructMappingFromAlign(fromStart, fromEnd,
198 MappingType.NucleotideToNucleotide);
202 acf.addMap(seq, mappedSequence, mapping);
203 align.addCodonFrame(acf);
205 } catch (NumberFormatException nfe)
207 System.err.println("Invalid start or end in Target " + target);
211 SequenceFeature sf = buildSequenceFeature(gffColumns, attributes);
216 * Returns the target sequence id extracted from the GFF name/value pairs.
217 * Default (standard behaviour) is the first token for "Target". This may be
218 * overridden where tools report this in a non-standard way.
221 * first token of a "Target" value from GFF column 9, typically
224 * a map with all parsed column 9 attributes
227 @SuppressWarnings("unused")
228 protected String findTargetId(String target, Map<String, List<String>> set)
234 * Processes one GFF 'protein_match'; fields of interest are
236 * <li>feature group - the database reporting a match e.g. Pfam</li>
237 * <li>Name - the matched entry's accession id in the database</li>
238 * <li>ID - a sequence identifier for the matched region (which may be
239 * appended as FASTA in the GFF file)</li>
243 * parsed GFF column 9 key/value(s)
245 * the sequence the GFF feature is on
247 * the sequence feature holding GFF data
249 * the alignment the sequence belongs to, where any new mappings
252 * a list of new 'virtual sequences' generated while parsing GFF
253 * @param relaxedIdMatching
254 * if true allow fuzzy search for a matching target sequence
255 * @return the (real or virtual) sequence(s) mapped to by this match
256 * @throws IOException
258 protected SequenceFeature processProteinMatch(
259 Map<String, List<String>> set, SequenceI seq,
260 String[] gffColumns, AlignmentI align, List<SequenceI> newseqs,
261 boolean relaxedIdMatching)
263 // This is currently tailored to InterProScan GFF output:
264 // ID holds the ID of the matched sequence, Target references the
265 // query sequence; this looks wrong, as ID should just be the GFF internal
266 // ID of the GFF feature, while Target would normally reference the matched
268 // TODO refactor as needed if other protein-protein GFF varies
270 SequenceFeature sf = buildSequenceFeature(gffColumns, set);
273 * locate the mapped sequence in the alignment, or as a
274 * (new or existing) virtual sequence in the newseqs list
276 List<String> targets = set.get(TARGET);
279 for (String target : targets)
282 SequenceI mappedSequence1 = findSequence(findTargetId(target, set), align,
283 newseqs, relaxedIdMatching);
284 SequenceI mappedSequence = mappedSequence1;
285 if (mappedSequence == null)
291 * give the mapped sequence a copy of the sequence feature, with
292 * start/end range adjusted
294 SequenceFeature sf2 = new SequenceFeature(sf);
296 int sequenceFeatureLength = 1 + sf.getEnd() - sf.getBegin();
297 sf2.setEnd(sequenceFeatureLength);
298 mappedSequence.addSequenceFeature(sf2);
301 * add a property to the mapped sequence so that it can eventually be
302 * renamed with its qualified accession id; renaming has to wait until
303 * all sequence reference resolution is complete
305 String accessionId = StringUtils.listToDelimitedString(
307 if (accessionId.length() > 0)
309 String database = sf.getType(); // TODO InterProScan only??
310 String qualifiedAccId = database + "|" + accessionId;
311 sf2.setValue(RENAME_TOKEN, qualifiedAccId);
315 * get any existing mapping for these sequences (or start one),
316 * and add this mapped range
318 AlignedCodonFrame alco = getMapping(align, seq, mappedSequence);
319 int[] from = new int[] { sf.getBegin(), sf.getEnd() };
320 int[] to = new int[] { 1, sequenceFeatureLength };
321 MapList mapping = new MapList(from, to, 1, 1);
323 alco.addMap(seq, mappedSequence, mapping);
324 align.addCodonFrame(alco);
332 * Return '=' as the name-value separator used in column 9 attributes.
335 protected char getNameValueSeparator()
341 * Modifies the default SequenceFeature in order to set the Target sequence id
345 protected SequenceFeature buildSequenceFeature(String[] gff,
346 Map<String, List<String>> attributes)
348 SequenceFeature sf = super.buildSequenceFeature(gff, attributes);
349 String desc = getDescription(sf, attributes);
352 sf.setDescription(desc);
358 * Apply heuristic rules to try to get the most useful feature description
364 protected String getDescription(SequenceFeature sf,
365 Map<String, List<String>> attributes)
368 String target = (String) sf.getValue(TARGET);
371 desc = target.split(" ")[0];
374 if (SequenceOntology.getInstance().isSequenceVariant(sf.getType()))
377 * Ensembl returns dna variants as 'alleles'
379 desc = StringUtils.listToDelimitedString(
380 attributes.get("alleles"), ",");