void
SVScorer::
getBreakendMaxMappedDepthAndMQ0(
const bool isMaxDepth,
const double cutoffDepth,
const SVBreakend& bp,
unsigned& maxDepth,
float& MQ0Frac)
{
/// define a new interval -/+ 50 bases around the center pos
/// of the breakpoint
static const pos_t regionSize(50);
maxDepth=0;
MQ0Frac=0;
unsigned totalReads(0);
unsigned totalMQ0Reads(0);
const pos_t centerPos(bp.interval.range.center_pos());
const known_pos_range2 searchRange(std::max((centerPos-regionSize),0), (centerPos+regionSize));
if (searchRange.size() == 0) return;
std::vector<unsigned> depth(searchRange.size(),0);
bool isCutoff(false);
bool isNormalFound(false);
const unsigned bamCount(_bamStreams.size());
for (unsigned bamIndex(0); bamIndex < bamCount; ++bamIndex)
{
if (_isAlignmentTumor[bamIndex]) continue;
isNormalFound=true;
bam_streamer& bamStream(*_bamStreams[bamIndex]);
// set bam stream to new search interval:
bamStream.set_new_region(bp.interval.tid, searchRange.begin_pos(), searchRange.end_pos());
while (bamStream.next())
{
const bam_record& bamRead(*(bamStream.get_record_ptr()));
// turn filtration down to mapped only to match depth estimate method:
if (bamRead.is_unmapped()) continue;
const pos_t refPos(bamRead.pos()-1);
if (refPos >= searchRange.end_pos()) break;
addReadToDepthEst(bamRead,searchRange.begin_pos(),depth);
totalReads++;
if (0 == bamRead.map_qual()) totalMQ0Reads++;
if (isMaxDepth)
{
const pos_t depthOffset(refPos-searchRange.begin_pos());
if (depthOffset>=0)
{
if (depth[depthOffset] > cutoffDepth)
{
isCutoff=true;
break;
}
}
}
}
if (isCutoff) break;
}
assert(isNormalFound);
maxDepth = *(std::max_element(depth.begin(),depth.end()));
if (totalReads>=10)
{
MQ0Frac = static_cast<float>(totalMQ0Reads)/static_cast<float>(totalReads);
}
}
/// 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()));
}
}
}