void
SVScorePairRefProcessor::
processClearedRecord(
const bam_record& bamRead)
{
using namespace illumina::common;
assert(bamParams.isSet);
const pos_t refPos(bamRead.pos()-1);
if (! bamParams.interval.range.is_pos_intersect(refPos)) return;
const bool isLargeInsert(isLargeInsertSV(sv));
#ifdef DEBUG_MEGAPAIR
log_os << __FUNCTION__ << ": read: " << bamRead << "\n";
#endif
/// check if fragment is too big or too small:
const int templateSize(std::abs(bamRead.template_size()));
if (templateSize < bamParams.minFrag) return;
if (templateSize > bamParams.maxFrag) return;
// count only from the down stream reads
const bool isFirstBamRead(isFirstRead(bamRead));
// get fragment range:
pos_t fragBeginRefPos(refPos);
if (! isFirstBamRead)
{
fragBeginRefPos=bamRead.mate_pos()-1;
}
const pos_t fragEndRefPos(fragBeginRefPos+templateSize);
if (fragBeginRefPos > fragEndRefPos)
{
std::ostringstream oss;
oss << "ERROR: Failed to parse fragment range from bam record. Frag begin,end: " << fragBeginRefPos << " " << fragEndRefPos << " bamRecord: " << bamRead << "\n";
BOOST_THROW_EXCEPTION(LogicException(oss.str()));
}
{
const pos_t fragOverlap(std::min((1+svParams.centerPos-fragBeginRefPos), (fragEndRefPos-svParams.centerPos)));
#ifdef DEBUG_MEGAPAIR
log_os << __FUNCTION__ << ": frag begin/end/overlap: " << fragBeginRefPos << " " << fragEndRefPos << " " << fragOverlap << "\n";
#endif
if (fragOverlap < pairOpt.minFragSupport) return;
}
SVFragmentEvidence& fragment(evidence.getSampleEvidence(bamParams.bamIndex)[bamRead.qname()]);
static const bool isShadow(false);
SVFragmentEvidenceRead& evRead(fragment.getRead(bamRead.is_first()));
setReadEvidence(svParams.minMapQ, svParams.minTier2MapQ, bamRead, isShadow, evRead);
setAlleleFrag(*bamParams.fragDistroPtr, templateSize, fragment.ref.getBp(isBp1),isLargeInsert);
}
static
bool
isGoodShadow(
const bam_record& bamRead,
const std::string& lastQname)
{
#ifdef DEBUG_IS_SHADOW
static const std::string logtag("isGoodShadow");
#endif
if (! bamRead.is_paired()) return false;
if (bamRead.isNonStrictSupplement()) return false;
// sanity check that this is a shadow read:
if (!bamRead.is_unmapped()) return false;
if (bamRead.is_mate_unmapped()) return false;
static const unsigned minAvgQualShadow = 25;
if (get_avg_quality(bamRead) < minAvgQualShadow)
{
return false;
}
if (strcmp(bamRead.qname(),lastQname.c_str()) != 0)
{
// something went wrong here, shadows should have their singleton partner
// preceding them in the BAM file.
#ifdef DEBUG_IS_SHADOW
log_os << logtag << " ERROR: Shadow without matching singleton : " << bamRead.qname() << " vs " << lastQname << std::endl;
#endif
return false;
}
#ifdef DEBUG_IS_SHADOW
log_os << logtag << " Found shadow!\n";
<< logtag << " this mapq = " << ((unsigned int)bamRead.map_qual()) << std::endl;
void
SVCandidateSetSequenceFragmentSampleGroup::
add(
const bam_header_info& bamHeader,
const bam_record& bamRead,
const bool isExpectRepeat,
const bool isSourcedFromGraphEdgeNode1,
const bool isSubMapped)
{
using namespace illumina::common;
#ifdef DEBUG_SVDATA
log_os << "SVDataGroup adding: " << bamRead << "\n";
#endif
SVCandidateSetSequenceFragment* fragPtr(getSequenceFragment(bamRead.qname()));
if (nullptr == fragPtr) return;
SVCandidateSetSequenceFragment& fragment(*fragPtr);
SVCandidateSetRead* targetReadPtr(nullptr);
if (2 == bamRead.read_no())
{
if (bamRead.isNonStrictSupplement())
{
fragment.read2Supplemental.emplace_back();
targetReadPtr = (&(fragment.read2Supplemental.back()));
}
else
{
targetReadPtr = (&(fragment.read2));
}
}
else
{
if (bamRead.isNonStrictSupplement())
{
fragment.read1Supplemental.emplace_back();
targetReadPtr = (&(fragment.read1Supplemental.back()));
}
else
{
targetReadPtr = (&(fragment.read1));
}
}
SVCandidateSetRead& targetRead(*targetReadPtr);
if (targetRead.isSet())
{
if (isExpectRepeat) return;
std::ostringstream oss;
oss << "Unexpected alignment name collision. Source: '" << dataSourceName << "'\n"
<< "\tExisting read: ";
summarizeAlignmentRecord(bamHeader, targetRead.bamrec, oss);
oss << "\n"
<< "\tNew read: ";
summarizeAlignmentRecord(bamHeader, bamRead, oss);
oss << "\n";
BOOST_THROW_EXCEPTION(GeneralException(oss.str()));
}
targetRead.bamrec = bamRead;
targetRead.isSourcedFromGraphEdgeNode1 = isSourcedFromGraphEdgeNode1;
targetRead.isSubMapped = isSubMapped;
targetRead.readIndex = (isSubMapped ? _subMappedReadIndex : _mappedReadIndex);
}
/// scan read record (and optionally its mate record) for SV evidence.
//
/// note that estimation is improved by the mate record (because we have the mate cigar string in this case)
///
static
void
getReadBreakendsImpl(
const ReadScannerOptions& opt,
const ReadScannerDerivOptions& dopt,
const SVLocusScanner::CachedReadGroupStats& rstats,
const bam_record& localRead,
const bam_record* remoteReadPtr,
const bam_header_info& bamHeader,
const reference_contig_segment& localRefSeq,
const reference_contig_segment* remoteRefSeqPtr,
std::vector<SVObservation>& candidates,
known_pos_range2& localEvidenceRange)
{
using namespace illumina::common;
#ifdef DEBUG_SCANNER
log_os << __FUNCTION__ << ": Starting read: " << localRead.qname() << "\n";
#endif
const chromMap_t& chromToIndex(bamHeader.chrom_to_index);
candidates.clear();
/// get some basic derived information from the bam_record:
const SimpleAlignment localAlign(getAlignment(localRead));
try
{
getSingleReadSVCandidates(opt, dopt, localRead, localAlign, chromToIndex,
localRefSeq, candidates);
// run the same check on the read's mate if we have access to it
if (nullptr != remoteReadPtr)
{
const bam_record& remoteRead(*remoteReadPtr);
const SimpleAlignment remoteAlign(getAlignment(remoteRead));
if (nullptr == remoteRefSeqPtr)
{
static const char msg[] = "ERROR: remoteRefSeqPtr cannot be null";
BOOST_THROW_EXCEPTION(LogicException(msg));
}
getSingleReadSVCandidates(opt, dopt, remoteRead, remoteAlign,
chromToIndex, (*remoteRefSeqPtr),
candidates);
}
// process shadows:
//getSVCandidatesFromShadow(opt, rstats, localRead, localAlign,remoteReadPtr,candidates);
// - process anomalous read pairs:
getSVCandidatesFromPair(opt, dopt, rstats, localRead, localAlign, remoteReadPtr,
candidates);
}
catch (...)
{
std::cerr << "ERROR: Exception caught while processing ";
if (nullptr == remoteReadPtr)
{
std::cerr << "single read record:\n"
<< '\t' << localRead << "\n";
}
else
{
std::cerr << " read pair records:\n"
<< '\t' << localRead << "\n"
<< '\t' << (*remoteReadPtr) << "\n";
}
throw;
}
#ifdef DEBUG_SCANNER
log_os << __FUNCTION__ << ": post-pair candidate_size: " << candidates.size() << "\n";
#endif
// update localEvidence range:
// note this is only used if candidates were added, so there's no harm in setting it every time:
const unsigned localRefLength(apath_ref_length(localAlign.path));
const pos_t startRefPos(localRead.pos()-1);
const pos_t endRefPos(startRefPos+localRefLength);
localEvidenceRange.set_range(startRefPos,endRefPos);
const int maxTid(chromToIndex.size());
/// final chance to QC candidate set:
///
for (const SVCandidate& sv : candidates)
{
bool isInvalidTid(false);
if ((sv.bp1.interval.tid < 0) || (sv.bp1.interval.tid >= maxTid))
{
isInvalidTid=true;
}
else if (sv.bp2.state != SVBreakendState::UNKNOWN)
{
if ((sv.bp2.interval.tid < 0) || (sv.bp2.interval.tid >= maxTid))
{
isInvalidTid=true;
}
}
bool isInvalidPos(false);
if (! isInvalidTid)
{
// note in the 'off-chromosome edge' test below we check for cases which are obviously way off
// the edge, but allow for a bit of over-edge mistakes to occur for the circular chromosomes
//
static const int offEdgePad(500);
const pos_t tid1Length(bamHeader.chrom_data[sv.bp1.interval.tid].length);
if ((sv.bp1.interval.range.end_pos() <= -offEdgePad) || (sv.bp1.interval.range.begin_pos() >= (tid1Length+offEdgePad)))
{
isInvalidPos=true;
}
else if (sv.bp2.state != SVBreakendState::UNKNOWN)
{
const pos_t tid2Length(bamHeader.chrom_data[sv.bp2.interval.tid].length);
if ((sv.bp2.interval.range.end_pos() <= -offEdgePad) || (sv.bp2.interval.range.begin_pos() >= (tid2Length+offEdgePad)))
{
isInvalidPos=true;
}
}
}
if (isInvalidTid || isInvalidPos)
{
std::ostringstream oss;
if (isInvalidTid)
{
oss << "SVbreakend has unknown or invalid chromosome id in candidate sv.\n";
}
else
{
oss << "Cannot interpret BAM record: candidate SV breakend from BAM record is off chromosome edge.\n";
}
oss << "\tlocal_bam_record: " << localRead << "\n"
<< "\tremote_bam record: ";
if (NULL==remoteReadPtr)
{
oss << "NONE";
}
else
{
oss << (*remoteReadPtr);
}
oss << "\n"
<< "\tSVCandidate: " << sv << "\n";
BOOST_THROW_EXCEPTION(LogicException(oss.str()));
}
}
}
/// get SV candidates from shadow/singleton pairs
/// look for singletons, create candidateSV around conf. interval of shadow position
/// cache singletons? might be needed to remove poor quality shadows.
/// should be able to re-use code, follow soft-clipping example.
static
void
getSVCandidatesFromShadow(
const ReadScannerOptions& opt,
const SVLocusScanner::CachedReadGroupStats& rstats,
const bam_record& localRead,
const SimpleAlignment& localAlign,
const bam_record* remoteReadPtr,
TrackedCandidates& candidates)
{
using namespace SVEvidenceType;
static const index_t svSource(SHADOW);
static const bool isComplex(true);
pos_t singletonGenomePos(0);
int targetId(0);
if (NULL == remoteReadPtr)
{
if (!localRead.is_unmapped()) return;
// need to take care of this case
// need to rely on cached mapq and qname
return;
if (!isGoodShadow(localRead,lastMapq,lastQname,opt.minSingletonMapqGraph))
{
return;
}
singletonGenomePos = localAlign.pos;
targetId = localRead.target_id();
}
else
{
// have both reads, straightforward from here
const bam_record& remoteRead(*remoteReadPtr);
const SimpleAlignment remoteAlign(remoteRead);
if (localRead.is_mate_unmapped())
{
// remote read is shadow candidate
if (!isGoodShadow(remoteRead,localRead.map_qual(),localRead.qname(),opt.minSingletonMapqGraph))
{
return;
}
singletonGenomePos = localAlign.pos;
targetId = remoteRead.target_id();
}
else if (localRead.is_unmapped())
{
// local is shadow candidate
if (!isGoodShadow(localRead,remoteRead.map_qual(),remoteRead.qname(),opt.minSingletonMapqGraph))
{
return;
}
singletonGenomePos = remoteAlign.pos;
targetId = localRead.target_id();
}
else
{
// none unmapped, skip this one
return;
}
}
const pos_t properPairRangeOffset = static_cast<pos_t>(rstats.properPair.min + (rstats.properPair.max-rstats.properPair.min)/2);
const pos_t shadowGenomePos = singletonGenomePos + properPairRangeOffset;
candidates.push_back(GetSplitSVCandidate(opt,targetId,shadowGenomePos,shadowGenomePos, svSource, isComplex));
}
