/**
* Retrieve the weight for this distribution.
*
* <p>Performs the standard munge to handle ambiguity symbols. The actual weights for each atomic
* symbol should be calculated by the getWeightImpl functions.
*
* @param sym the Symbol to find the probability of
* @return the probability that one of the symbols matching amb was emitted
* @throws IllegalSymbolException if for any reason the symbols within amb are not recognized by
* this state
*/
public final double getWeight(Symbol sym) throws IllegalSymbolException {
if (sym instanceof AtomicSymbol) {
return getWeightImpl((AtomicSymbol) sym);
} else {
Alphabet ambA = sym.getMatches();
if (((FiniteAlphabet) ambA).size() == 0) { // a gap
getAlphabet().validate(sym);
double totalWeight = 0.0;
for (Iterator i = ((FiniteAlphabet) getAlphabet()).iterator(); i.hasNext(); ) {
Symbol s = (Symbol) i.next();
totalWeight += getWeight(s);
}
return 1.0 - totalWeight;
}
if (ambA instanceof FiniteAlphabet) {
FiniteAlphabet fa = (FiniteAlphabet) ambA;
double sum = 0.0;
for (Iterator i = fa.iterator(); i.hasNext(); ) {
Object obj = i.next();
if (!(obj instanceof AtomicSymbol)) {
throw new BioError("Assertion Failure: Not an instance of AtomicSymbol: " + obj);
}
AtomicSymbol as = (AtomicSymbol) obj;
sum += getWeightImpl(as);
}
return sum;
} else {
throw new IllegalSymbolException(
"Can't find weight for infinite set of symbols matched by " + sym.getName());
}
}
}
private void doSetWeight(Symbol sym, double weight)
throws IllegalSymbolException, ChangeVetoException {
if (sym instanceof AtomicSymbol) {
setWeightImpl((AtomicSymbol) sym, weight);
} else {
// need to divide the weight up amongst the atomic symbols according
// to the null model
FiniteAlphabet fa = (FiniteAlphabet) sym.getMatches();
double totalNullWeight = this.getNullModel().getWeight(sym);
for (Iterator si = fa.iterator(); si.hasNext(); ) {
AtomicSymbol as = (AtomicSymbol) si.next();
double symNullWeight = this.getNullModel().getWeight(as);
setWeightImpl(as, weight * symNullWeight / totalNullWeight);
}
}
}
/**
* <code>makeSubHit</code> creates a new sub-hit.
*
* @return a <code>SeqSimilaritySearchSubHit</code>.
* @exception BioException if an error occurs.
*/
private SeqSimilaritySearchSubHit makeSubHit() throws BioException {
// Try to get a valid TokenParser
if (tokenParser == null) {
String identifier;
// Try explicit sequence type first
if (subHitData.containsKey("subjectSequenceType"))
identifier = (String) subHitData.get("subjectSequenceType");
// Otherwise try to resolve from the program name (only
// works for Blast)
else if (resultPreAnnotation.containsKey("program"))
identifier = (String) resultPreAnnotation.get("program");
else throw new BioException("Failed to determine sequence type");
FiniteAlphabet alpha = AlphabetResolver.resolveAlphabet(identifier);
tokenParser = alpha.getTokenization("token");
}
// BLASTP output has the strands set null (protein sequences)
Strand qStrand = null;
Strand sStrand = null;
// Override where an explicit strand is given (FASTA DNA,
// BLASTN)
if (subHitData.containsKey("queryStrand"))
if (subHitData.get("queryStrand").equals("plus")) qStrand = StrandedFeature.POSITIVE;
else qStrand = StrandedFeature.NEGATIVE;
if (subHitData.containsKey("subjectStrand"))
if (subHitData.get("subjectStrand").equals("plus")) sStrand = StrandedFeature.POSITIVE;
else sStrand = StrandedFeature.NEGATIVE;
// Override where a frame is given as this contains strand
// information (BLASTX for query, TBLASTN for hit, TBLASTX for
// both)
if (subHitData.containsKey("queryFrame"))
if (((String) subHitData.get("queryFrame")).startsWith("plus"))
qStrand = StrandedFeature.POSITIVE;
else qStrand = StrandedFeature.NEGATIVE;
if (subHitData.containsKey("subjectFrame"))
if (((String) subHitData.get("subjectFrame")).startsWith("plus"))
sStrand = StrandedFeature.POSITIVE;
else sStrand = StrandedFeature.NEGATIVE;
// Get start/end
int qStart = Integer.parseInt((String) subHitData.get("querySequenceStart"));
int qEnd = Integer.parseInt((String) subHitData.get("querySequenceEnd"));
int sStart = Integer.parseInt((String) subHitData.get("subjectSequenceStart"));
int sEnd = Integer.parseInt((String) subHitData.get("subjectSequenceEnd"));
// The start/end coordinates from BioJava XML don't follow the
// BioJava paradigm of start < end, with orientation given by
// the strand property. Rather, they present start/end as
// displayed in BLAST output, with the coordinates being
// inverted on the reverse strand. We account for this here.
if (qStrand == StrandedFeature.NEGATIVE) {
int swap = qStart;
qStart = qEnd;
qEnd = swap;
}
if (sStrand == StrandedFeature.NEGATIVE) {
int swap = sStart;
sStart = sEnd;
sEnd = swap;
}
// Get scores
double sc = Double.NaN;
double ev = Double.NaN;
double pv = Double.NaN;
if (subHitData.containsKey("score")) sc = Double.parseDouble((String) subHitData.get("score"));
if (subHitData.containsKey("expectValue")) {
String val = (String) subHitData.get("expectValue");
// Blast sometimes uses invalid formatting such as 'e-156'
// rather than '1e-156'
if (val.startsWith("e")) ev = Double.parseDouble("1" + val);
else ev = Double.parseDouble(val);
}
if (subHitData.containsKey("pValue"))
pv = Double.parseDouble((String) subHitData.get("pValue"));
Map labelMap = new SmallMap();
// Note that the following is removing the raw sequences
StringBuffer tokenBuffer = new StringBuffer(1024);
tokenBuffer.append((String) subHitData.remove("querySequence"));
labelMap.put(
SeqSimilaritySearchSubHit.QUERY_LABEL,
new SimpleSymbolList(tokenParser, tokenBuffer.substring(0)));
tokenBuffer = new StringBuffer(1024);
tokenBuffer.append((String) subHitData.remove("subjectSequence"));
labelMap.put(
hitData.get("subjectId"), new SimpleSymbolList(tokenParser, tokenBuffer.substring(0)));
return new SimpleSeqSimilaritySearchSubHit(
sc,
ev,
pv,
qStart,
qEnd,
qStrand,
sStart,
sEnd,
sStrand,
new SimpleAlignment(labelMap),
AnnotationFactory.makeAnnotation(subHitData));
}
