/**
* Finds the adaptor boundary around the read and returns the first base inside the adaptor that
* is closest to the read boundary. If the read is in the positive strand, this is the first base
* after the end of the fragment (Picard calls it 'insert'), if the read is in the negative
* strand, this is the first base before the beginning of the fragment.
*
* <p>There are two cases we need to treat here:
*
* <p>1) Our read is in the reverse strand :
*
* <p><----------------------| * |--------------------->
*
* <p>in these cases, the adaptor boundary is at the mate start (minus one)
*
* <p>2) Our read is in the forward strand :
*
* <p>|----------------------> * <----------------------|
*
* <p>in these cases the adaptor boundary is at the start of the read plus the inferred insert
* size (plus one)
*
* @param read the read being tested for the adaptor boundary
* @return the reference coordinate for the adaptor boundary (effectively the first base IN the
* adaptor, closest to the read. NULL if the read is unmapped or the mate is mapped to another
* contig.
*/
public static Integer getAdaptorBoundary(final SAMRecord read) {
final int MAXIMUM_ADAPTOR_LENGTH = 8;
final int insertSize =
Math.abs(
read
.getInferredInsertSize()); // the inferred insert size can be negative if the mate
// is mapped before the read (so we take the absolute
// value)
if (insertSize == 0
|| read
.getReadUnmappedFlag()) // no adaptors in reads with mates in another chromosome or
// unmapped pairs
return null;
Integer
adaptorBoundary; // the reference coordinate for the adaptor boundary (effectively the first
// base IN the adaptor, closest to the read)
if (read.getReadNegativeStrandFlag())
adaptorBoundary = read.getMateAlignmentStart() - 1; // case 1 (see header)
else adaptorBoundary = read.getAlignmentStart() + insertSize + 1; // case 2 (see header)
if ((adaptorBoundary < read.getAlignmentStart() - MAXIMUM_ADAPTOR_LENGTH)
|| (adaptorBoundary > read.getAlignmentEnd() + MAXIMUM_ADAPTOR_LENGTH))
adaptorBoundary =
null; // we are being conservative by not allowing the adaptor boundary to go beyond what
// we belive is the maximum size of an adaptor
return adaptorBoundary;
}
private boolean passesFilter(
final SAMRecord record,
final String sequence,
final int startPos,
final int endPos,
final boolean contained) {
if (record == null) {
return false;
}
if (!safeEquals(record.getReferenceName(), sequence)) {
return false;
}
final int alignmentStart = record.getAlignmentStart();
int alignmentEnd = record.getAlignmentEnd();
if (alignmentStart <= 0) {
assertTrue(record.getReadUnmappedFlag());
return false;
}
if (alignmentEnd <= 0) {
// For indexing-only records, treat as single base alignment.
assertTrue(record.getReadUnmappedFlag());
alignmentEnd = alignmentStart;
}
if (contained) {
if (startPos != 0 && alignmentStart < startPos) {
return false;
}
if (endPos != 0 && alignmentEnd > endPos) {
return false;
}
} else {
if (startPos != 0 && alignmentEnd < startPos) {
return false;
}
if (endPos != 0 && alignmentStart > endPos) {
return false;
}
}
return true;
}
/**
* Returns the duplicate score computed from the given fragment. value should be capped by
* Short.MAX_VALUE/2 since the score from two reads will be added and an overflow will be
*
* <p>If true is given to assumeMateCigar, then any score that can use the mate cigar to compute
* the mate's score will return the score computed on both ends.
*/
public static short computeDuplicateScore(
final SAMRecord record,
final ScoringStrategy scoringStrategy,
final boolean assumeMateCigar) {
Short storedScore = (Short) record.getTransientAttribute(Attr.DuplicateScore);
if (storedScore == null) {
short score = 0;
switch (scoringStrategy) {
case SUM_OF_BASE_QUALITIES:
// two (very) long reads worth of high-quality bases can go over Short.MAX_VALUE/2
// and risk overflow.
score += (short) Math.min(getSumOfBaseQualities(record), Short.MAX_VALUE / 2);
break;
case TOTAL_MAPPED_REFERENCE_LENGTH:
if (!record.getReadUnmappedFlag()) {
// no need to remember the score since this scoring mechanism is symmetric
score = (short) Math.min(record.getCigar().getReferenceLength(), Short.MAX_VALUE / 2);
}
if (assumeMateCigar && record.getReadPairedFlag() && !record.getMateUnmappedFlag()) {
score +=
(short)
Math.min(
SAMUtils.getMateCigar(record).getReferenceLength(), Short.MAX_VALUE / 2);
}
break;
// The RANDOM score gives the same score to both reads so that they get filtered together.
// it's not critical do use the readName since the scores from both ends get added, but it
// seem
// to be clearer this way.
case RANDOM:
// start with a random number between Short.MIN_VALUE/4 and Short.MAX_VALUE/4
score += (short) (hasher.hashUnencodedChars(record.getReadName()) & 0b11_1111_1111_1111);
// subtract Short.MIN_VALUE/4 from it to end up with a number between
// 0 and Short.MAX_VALUE/2. This number can be then discounted in case the read is
// not passing filters. We need to stay far from overflow so that when we add the two
// scores from the two read mates we do not overflow since that could cause us to chose a
// failing read-pair instead of a passing one.
score -= Short.MIN_VALUE / 4;
}
// make sure that filter-failing records are heavily discounted. (the discount can happen
// twice, once
// for each mate, so need to make sure we do not subtract more than Short.MIN_VALUE overall.)
score += record.getReadFailsVendorQualityCheckFlag() ? (short) (Short.MIN_VALUE / 2) : 0;
storedScore = score;
record.setTransientAttribute(Attr.DuplicateScore, storedScore);
}
return storedScore;
}
private void collectReadData(final SAMRecord record, final ReferenceSequence ref) {
metrics.TOTAL_READS++;
readLengthHistogram.increment(record.getReadBases().length);
if (!record.getReadFailsVendorQualityCheckFlag()) {
metrics.PF_READS++;
if (isNoiseRead(record)) metrics.PF_NOISE_READS++;
if (record.getReadUnmappedFlag()) {
// If the read is unmapped see if it's adapter sequence
final byte[] readBases = record.getReadBases();
if (!(record instanceof BAMRecord)) StringUtil.toUpperCase(readBases);
if (isAdapterSequence(readBases)) {
this.adapterReads++;
}
} else if (doRefMetrics) {
metrics.PF_READS_ALIGNED++;
if (!record.getReadNegativeStrandFlag()) numPositiveStrand++;
if (record.getReadPairedFlag() && !record.getMateUnmappedFlag()) {
metrics.READS_ALIGNED_IN_PAIRS++;
// Check that both ends have mapq > minimum
final Integer mateMq = record.getIntegerAttribute("MQ");
if (mateMq == null
|| mateMq >= MAPPING_QUALITY_THRESOLD
&& record.getMappingQuality() >= MAPPING_QUALITY_THRESOLD) {
++this.chimerasDenominator;
// With both reads mapped we can see if this pair is chimeric
if (Math.abs(record.getInferredInsertSize()) > maxInsertSize
|| !record.getReferenceIndex().equals(record.getMateReferenceIndex())) {
++this.chimeras;
}
}
}
}
}
}
private void collectQualityData(final SAMRecord record, final ReferenceSequence reference) {
// If the read isnt an aligned PF read then look at the read for no-calls
if (record.getReadUnmappedFlag()
|| record.getReadFailsVendorQualityCheckFlag()
|| !doRefMetrics) {
final byte[] readBases = record.getReadBases();
for (int i = 0; i < readBases.length; i++) {
if (SequenceUtil.isNoCall(readBases[i])) {
badCycleHistogram.increment(
CoordMath.getCycle(record.getReadNegativeStrandFlag(), readBases.length, i));
}
}
} else if (!record.getReadFailsVendorQualityCheckFlag()) {
final boolean highQualityMapping = isHighQualityMapping(record);
if (highQualityMapping) metrics.PF_HQ_ALIGNED_READS++;
final byte[] readBases = record.getReadBases();
final byte[] refBases = reference.getBases();
final byte[] qualities = record.getBaseQualities();
final int refLength = refBases.length;
long mismatchCount = 0;
long hqMismatchCount = 0;
for (final AlignmentBlock alignmentBlock : record.getAlignmentBlocks()) {
final int readIndex = alignmentBlock.getReadStart() - 1;
final int refIndex = alignmentBlock.getReferenceStart() - 1;
final int length = alignmentBlock.getLength();
for (int i = 0; i < length && refIndex + i < refLength; ++i) {
final int readBaseIndex = readIndex + i;
boolean mismatch =
!SequenceUtil.basesEqual(readBases[readBaseIndex], refBases[refIndex + i]);
boolean bisulfiteBase = false;
if (mismatch && isBisulfiteSequenced) {
if ((record.getReadNegativeStrandFlag()
&& (refBases[refIndex + i] == 'G' || refBases[refIndex + i] == 'g')
&& (readBases[readBaseIndex] == 'A' || readBases[readBaseIndex] == 'a'))
|| ((!record.getReadNegativeStrandFlag())
&& (refBases[refIndex + i] == 'C' || refBases[refIndex + i] == 'c')
&& (readBases[readBaseIndex] == 'T')
|| readBases[readBaseIndex] == 't')) {
bisulfiteBase = true;
mismatch = false;
}
}
if (mismatch) mismatchCount++;
metrics.PF_ALIGNED_BASES++;
if (!bisulfiteBase) nonBisulfiteAlignedBases++;
if (highQualityMapping) {
metrics.PF_HQ_ALIGNED_BASES++;
if (!bisulfiteBase) hqNonBisulfiteAlignedBases++;
if (qualities[readBaseIndex] >= BASE_QUALITY_THRESHOLD)
metrics.PF_HQ_ALIGNED_Q20_BASES++;
if (mismatch) hqMismatchCount++;
}
if (mismatch || SequenceUtil.isNoCall(readBases[readBaseIndex])) {
badCycleHistogram.increment(
CoordMath.getCycle(record.getReadNegativeStrandFlag(), readBases.length, i));
}
}
}
mismatchHistogram.increment(mismatchCount);
hqMismatchHistogram.increment(hqMismatchCount);
// Add any insertions and/or deletions to the global count
for (final CigarElement elem : record.getCigar().getCigarElements()) {
final CigarOperator op = elem.getOperator();
if (op == CigarOperator.INSERTION || op == CigarOperator.DELETION) ++this.indels;
}
}
}
public void find_coverage(SAMResource sres) {
int start_base = sres.region.range.start;
int end_base = sres.region.range.end;
int coverage_len = (end_base - start_base) + 1;
int i, end, ref_i, read_i, len;
int[] coverage = new int[coverage_len];
Arrays.fill(coverage, 0);
WorkingFile wf = null;
if (outfile != null) {
try {
wf = new WorkingFile(outfile);
ps = wf.getPrintStream();
} catch (Exception e) {
System.err.println("I/O error: " + e); // debug
e.printStackTrace();
System.exit(1);
}
}
try {
//
// gather coverage info:
//
CloseableIterator<SAMRecord> iterator = sres.get_iterator();
int read_count = 0;
int ref_min = -1;
int ref_max = -1;
while (iterator.hasNext()) {
SAMRecord sr = iterator.next();
read_count++;
// System.err.println(sr.getReadName() + ": " + sr.getAlignmentStart() + "-" +
// sr.getAlignmentEnd()); // debug
if (sr.getReadUnmappedFlag()) continue;
if (sr.getDuplicateReadFlag()) {
if (verbose_mode)
System.err.println(
sr.getReadName()
+ "."
+ (sr.getReadNegativeStrandFlag() ? "R" : "F")
+ " ignoring, duplicate");
continue;
}
byte[] read = sr.getReadBases();
byte[] quals = sr.getBaseQualities();
for (AlignmentBlock ab : sr.getAlignmentBlocks()) {
len = ab.getLength();
read_i = ab.getReadStart() - 1;
ref_i = ab.getReferenceStart() - start_base;
if (ref_min == -1 || ref_i < ref_min) ref_min = ref_i;
for (i = read_i, end = read_i + len; i < end; i++, ref_i++) {
if (ref_i >= 0 && ref_i < coverage_len) {
if (quals[i] >= MIN_QUALITY) {
if (verbose_mode)
System.err.println(
sr.getReadName()
+ "."
+ (sr.getReadNegativeStrandFlag() ? "R" : "F")
+ " hit at "
+ (ref_i + start_base)
+ " as="
+ sr.getAlignmentStart()
+ " ae="
+ sr.getAlignmentEnd());
coverage[ref_i]++;
} else if (verbose_mode) {
System.err.println(
sr.getReadName()
+ "."
+ (sr.getReadNegativeStrandFlag() ? "R" : "F")
+ " qual_reject at "
+ (ref_i + start_base)
+ " as="
+ sr.getAlignmentStart()
+ " ae="
+ sr.getAlignmentEnd());
}
}
}
if (ref_max == -1 || ref_i > ref_max) ref_max = ref_i;
}
}
sres.close();
System.err.println(
"records:"
+ read_count
+ " ref_min:"
+ (ref_min + start_base)
+ " ref_max:"
+ (ref_max + start_base)); // debug
//
// report coverage info:
//
for (i = 0; i < coverage.length; i++) {
if (name != null) ps.print(name + ",");
ps.println((i + start_base) + "," + coverage[i]); // debug
}
if (wf != null) wf.finish();
} catch (Exception e) {
System.err.println("ERROR: " + e); // debug
e.printStackTrace();
}
}
