/**
* Use the given read to build a list of aligned read and reference base information.
*
* @param read The read with the alignment information
* @return read and reference information. For a read without an alignment or cigar units, null is
* returned.
*/
public static List<ReadBaseWithReference> extractReadBases(Read read) {
// Make sure this read has a valid alignment with Cigar Units
if (!read.hasAlignment() || (read.getAlignment().getCigarCount() == 0)) {
return null;
}
ImmutableList.Builder<ReadBaseWithReference> bases = ImmutableList.builder();
String readSeq = read.getAlignedSequence();
List<Integer> readQual = read.getAlignedQualityList();
String refSeq = UNINITIALIZED_REFERENCE_SEQUENCE;
int refPosAbsoluteOffset = 0;
int readOffset = 0;
for (CigarUnit unit : read.getAlignment().getCigarList()) {
switch (unit.getOperation()) {
case ALIGNMENT_MATCH:
case SEQUENCE_MISMATCH:
case SEQUENCE_MATCH:
for (int i = 0; i < unit.getOperationLength(); i++) {
String refBase = "";
if (unit.getOperation().equals(CigarUnit.Operation.SEQUENCE_MATCH)) {
refBase = readSeq.substring(readOffset, readOffset + 1);
} else if (!unit.getReferenceSequence().isEmpty()) {
// try to get the ref sequence from the Cigar unit
refBase = unit.getReferenceSequence().substring(i, i + 1);
} else {
// try to get the ref sequence by fully parsing the MD tag if not already cached
if (refSeq != null && refSeq.equals(UNINITIALIZED_REFERENCE_SEQUENCE)) {
refSeq =
com.google.cloud.genomics.utils.grpc.ReadUtils
.inferReferenceSequenceByParsingMdFlag(read);
}
if (refSeq != null) {
refBase = refSeq.substring(readOffset, readOffset + 1);
}
}
String name = read.getAlignment().getPosition().getReferenceName();
Matcher m = Pattern.compile("^(chr)?(X|Y|([12]?\\d))$").matcher(name);
if (m.matches()) {
name = m.group(m.groupCount() - 1);
}
Position refPosition =
Position.newBuilder()
.setReferenceName(name)
.setPosition(
read.getAlignment().getPosition().getPosition() + refPosAbsoluteOffset)
.build();
bases.add(
new ReadBaseWithReference(
new ReadBaseQuality(
readSeq.substring(readOffset, readOffset + 1), readQual.get(readOffset)),
refBase,
refPosition));
refPosAbsoluteOffset++;
readOffset++;
}
break;
case PAD: // padding (silent deletion from padded reference)
case CLIP_HARD: // hard clipping (clipped sequences NOT present in seq)
break;
case CLIP_SOFT: // soft clipping (clipped sequences present in SEQ)
case INSERT: // insertion to the reference
readOffset += unit.getOperationLength();
break;
case DELETE: // deletion from the reference
case SKIP: // intron (mRNA-to-genome alignment only)
refPosAbsoluteOffset += unit.getOperationLength();
break;
default:
throw new IllegalArgumentException("Illegal cigar code: " + unit.getOperation());
}
}
return bases.build();
}
@Override
public Boolean apply(Read r) {
return r.getProperPlacement();
}
@Override
public Boolean apply(Read r) {
return !r.getFailedVendorQualityChecks();
}
@Override
public Boolean apply(Read r) {
return !r.getDuplicateFragment();
}
@Override
public Boolean apply(Read r) {
return Pattern.compile("^(chr)?(X|Y|([12]?\\d))$")
.matcher(r.getAlignment().getPosition().getReferenceName())
.matches();
}
