private SnpEffEffect getMostSignificantEffect(List<SnpEffEffect> effects) {
SnpEffEffect mostSignificantEffect = null;
for (SnpEffEffect effect : effects) {
if (mostSignificantEffect == null || effect.isHigherImpactThan(mostSignificantEffect)) {
mostSignificantEffect = effect;
}
}
return mostSignificantEffect;
}
private List<SnpEffEffect> parseSnpEffRecord(VariantContext snpEffRecord) {
List<SnpEffEffect> parsedEffects = new ArrayList<SnpEffEffect>();
Object effectFieldValue = snpEffRecord.getAttribute(SNPEFF_INFO_FIELD_KEY);
if (effectFieldValue == null) {
return parsedEffects;
}
// The VCF codec stores multi-valued fields as a List<String>, and single-valued fields as a
// String.
// We can have either in the case of SnpEff, since there may be one or more than one effect in
// this record.
List<String> individualEffects;
if (effectFieldValue instanceof List) {
individualEffects = (List<String>) effectFieldValue;
} else {
individualEffects = Arrays.asList((String) effectFieldValue);
}
for (String effectString : individualEffects) {
String[] effectNameAndMetadata = effectString.split(SNPEFF_EFFECT_METADATA_DELIMITER);
if (effectNameAndMetadata.length != 2) {
logger.warn(
String.format(
"Malformed SnpEff effect field at %s:%d, skipping: %s",
snpEffRecord.getChr(), snpEffRecord.getStart(), effectString));
continue;
}
String effectName = effectNameAndMetadata[0];
String[] effectMetadata =
effectNameAndMetadata[1].split(SNPEFF_EFFECT_METADATA_SUBFIELD_DELIMITER, -1);
SnpEffEffect parsedEffect = new SnpEffEffect(effectName, effectMetadata);
if (parsedEffect.isWellFormed()) {
parsedEffects.add(parsedEffect);
} else {
logger.warn(
String.format(
"Skipping malformed SnpEff effect field at %s:%d. Error was: \"%s\". Field was: \"%s\"",
snpEffRecord.getChr(),
snpEffRecord.getStart(),
parsedEffect.getParseError(),
effectString));
}
}
return parsedEffects;
}
public boolean isHigherImpactThan(SnpEffEffect other) {
// If one effect is within a coding gene and the other is not, the effect that is
// within the coding gene has higher impact:
if (isCoding() && !other.isCoding()) {
return true;
} else if (!isCoding() && other.isCoding()) {
return false;
}
// Otherwise, both effects are either in or not in a coding gene, so we compare the impacts
// of the effects themselves. Effects with the same impact are tie-broken using the
// functional class of the effect:
if (impact.isHigherImpactThan(other.impact)) {
return true;
} else if (impact.isSameImpactAs(other.impact)) {
return functionalClass.isHigherPriorityThan(other.functionalClass);
}
return false;
}
@Override
public Map<String, Object> annotate(
final RefMetaDataTracker tracker,
final AnnotatorCompatible walker,
final ReferenceContext ref,
final Map<String, AlignmentContext> stratifiedContexts,
final VariantContext vc,
final Map<String, PerReadAlleleLikelihoodMap> stratifiedPerReadAlleleLikelihoodMap) {
// Can not annotate if failed initialization conditions
if (!canAnnotate) return null;
RodBinding<VariantContext> snpEffRodBinding = walker.getSnpEffRodBinding();
// Get only SnpEff records that start at this locus, not merely span it:
List<VariantContext> snpEffRecords = tracker.getValues(snpEffRodBinding, ref.getLocus());
// Within this set, look for a SnpEff record whose ref/alt alleles match the record to annotate.
// If there is more than one such record, we only need to pick the first one, since the
// biological
// effects will be the same across all such records:
VariantContext matchingRecord = getMatchingSnpEffRecord(snpEffRecords, vc);
if (matchingRecord == null) {
return null;
}
// Parse the SnpEff INFO field annotation from the matching record into individual effect
// objects:
List<SnpEffEffect> effects = parseSnpEffRecord(matchingRecord);
if (effects.isEmpty()) {
return null;
}
// Add only annotations for one of the most biologically-significant effects from this set:
SnpEffEffect mostSignificantEffect = getMostSignificantEffect(effects);
return mostSignificantEffect.getAnnotations();
}