void FragmentCollector::add(
const alignment::BamTemplate &bamTemplate,
const unsigned fragmentIndex,
const unsigned barcodeIdx)
{
ISAAC_ASSERT_MSG(2 >= bamTemplate.getFragmentCount(), "Expected paired or single-ended data");
const alignment::FragmentMetadata &fragment = bamTemplate.getFragmentMetadata(fragmentIndex);
ISAAC_THREAD_CERR_DEV_TRACE_CLUSTER_ID(fragment.getCluster().getId(), "FragmentCollector::add: " << fragment);
FragmentBuffer::IndexRecord &recordStart =
buffer_.initialize(fragment.getCluster().getId(), fragment.getReadIndex());
recordStart.fStrandPos_ = fragment.getFStrandReferencePosition();
storeBclAndCigar(fragment, recordStart);
if (2 == bamTemplate.getFragmentCount())
{
const alignment::FragmentMetadata &mate = bamTemplate.getMateFragmentMetadata(fragment);
if (fragment.isNoMatch())
{
ISAAC_ASSERT_MSG(mate.isNoMatch(), "If mate is not a no-match, fragment must be a shadow. fragment: "
<< fragment << " mate:" << mate);
recordStart.fragmentHeader() = io::FragmentHeader(bamTemplate, fragment, mate, barcodeIdx, 0);
}
else
{
const unsigned mateStorageBin = binIndexMap_.getBinIndex(mate.getFStrandReferencePosition());
recordStart.fragmentHeader() = io::FragmentHeader(bamTemplate, fragment, mate, barcodeIdx,
mateStorageBin);
}
}
else
{
recordStart.fragmentHeader() = io::FragmentHeader(bamTemplate, fragment, barcodeIdx);
}
}
/**
* \brief Removes entries designating the same alignment location and strand
*
* This is important not only to avoid repetitive processing but also to avoid good alignments
* scored poorly because they are present multiple times in the list.
*
* \param removeUnaligned if true, entries with !isAlignment() are removed
*
* \postcondition the fragmentList is ordered
* \postcondition The fragmentList contains unique alignments only
*/
void FragmentBuilder::consolidateDuplicateAlignments(
FragmentMetadataList &fragmentList,
const bool removeUnaligned,
std::size_t &perfectFound)
{
ISAAC_THREAD_CERR_DEV_TRACE_CLUSTER_ID(fragmentList[0].getCluster().getId(), "consolidateDuplicateFragments initial size: " << fragmentList.size());
perfectFound = 0;
// although initially matches arrive ordered by location, gapped alignment might have moved the
// start position of some.
std::sort(fragmentList.begin(), fragmentList.end());
FragmentMetadataList::iterator lastFragment = fragmentList.begin();
while(fragmentList.end() != lastFragment && removeUnaligned && !lastFragment->isAligned())
{
++lastFragment;
}
lastFragment = fragmentList.erase(fragmentList.begin(), lastFragment);
if (2 <= fragmentList.size())
{
ISAAC_THREAD_CERR_DEV_TRACE_CLUSTER_ID(fragmentList[0].getCluster().getId(), " keeping " << *lastFragment);
perfectFound += lastFragment->isPerfectMatch();
for (FragmentMetadataList::iterator currentFragment = lastFragment + 1; fragmentList.end() != currentFragment; ++currentFragment)
{
if (removeUnaligned && !currentFragment->isAligned())
{
ISAAC_THREAD_CERR_DEV_TRACE_CLUSTER_ID(fragmentList[0].getCluster().getId(), " ignoring " << *currentFragment);
}
else if (*lastFragment == *currentFragment)
{
ISAAC_THREAD_CERR_DEV_TRACE_CLUSTER_ID(fragmentList[0].getCluster().getId(), " merging " << *currentFragment);
lastFragment->mergeAnchors(*currentFragment);
ISAAC_THREAD_CERR_DEV_TRACE_CLUSTER_ID(fragmentList[0].getCluster().getId(), " merged " << *lastFragment);
}
else
{
ISAAC_THREAD_CERR_DEV_TRACE_CLUSTER_ID(fragmentList[0].getCluster().getId(), " keeping " << *currentFragment);
perfectFound += currentFragment->isPerfectMatch();
++lastFragment;
if (lastFragment != currentFragment)
{
*lastFragment = *currentFragment;
}
}
}
fragmentList.resize(1 + lastFragment - fragmentList.begin());
}
ISAAC_THREAD_CERR_DEV_TRACE_CLUSTER_ID(
fragmentList[0].getCluster().getId(),
"consolidateDuplicateFragments consolidated size: " << fragmentList.size() << " found " << perfectFound << " perfect alignments");
}
