// Classify the blocks in obList as irreducible, transitive or substrings. The irreducible blocks are
// put into pOBFinal. The remaining are discarded.
// Invariant: the blocks are ordered in descending order of the overlap size so that the longest overlap is first.
void OverlapAlgorithm::_processIrreducibleBlocksInexact(const BWT* pBWT, const BWT* pRevBWT,
OverlapBlockList& activeList,
OverlapBlockList* pOBFinal) const
{
if(activeList.empty())
return;
// The activeList contains all the blocks that are not yet right terminal
// Count the extensions in the top level (longest) blocks first
bool all_eliminated = false;
while(!activeList.empty() && !all_eliminated)
{
// The terminalBlock list contains all the blocks that became right-terminal
// in the current extension round.
OverlapBlockList terminalList;
OverlapBlockList potentialContainedList;
// Perform a single round of extension, any terminal blocks
// are moved to the terminated list
extendActiveBlocksRight(pBWT, pRevBWT, activeList, terminalList, potentialContainedList);
// Compare the blocks in the contained list against the other terminal and active blocks
// If they are a substring match to any of these, discard them
OverlapBlockList::iterator containedIter = potentialContainedList.begin();
for(; containedIter != potentialContainedList.end(); ++containedIter)
{
if(!isBlockSubstring(*containedIter, terminalList, m_errorRate) &&
!isBlockSubstring(*containedIter, activeList, m_errorRate))
{
// Not a substring, move to terminal list
terminalList.push_back(*containedIter);
//std::cout << "Contained block kept: " << containedIter->overlapLen << "\n";
}
else
{
//std::cout << "Contained block found and removed: " << containedIter->overlapLen << "\n";
}
}
// Using the terminated blocks, mark as eliminated any active blocks
// that form a valid overlap to the terminal block. These are transitive edges
// We do not compare two terminal blocks, we don't consider these overlaps to be
// transitive
OverlapBlockList::iterator terminalIter = terminalList.begin();
for(; terminalIter != terminalList.end(); ++terminalIter)
{
#ifdef DEBUGOVERLAP
std::cout << "[II] ***TLB of length " << terminalIter->overlapLen << " has ended\n";
#endif
all_eliminated = true;
OverlapBlockList::iterator activeIter = activeList.begin();
for(; activeIter != activeList.end(); ++activeIter)
{
if(activeIter->isEliminated)
continue; // skip previously marked blocks
// Two conditions must be met for a block to be transitive wrt terminal:
// 1) It must have a strictly shorter overlap than the terminal block
// 2) The error rate between the block and terminal must be less than the threshold
double inferredErrorRate = calculateBlockErrorRate(*terminalIter, *activeIter);
if(activeIter->overlapLen < terminalIter->overlapLen &&
isErrorRateAcceptable(inferredErrorRate, m_errorRate))
{
#ifdef DEBUGOVERLAP_2
std::cout << "Marking block of length " << activeIter->overlapLen << " as eliminated\n";
#endif
activeIter->isEliminated = true;
}
else
{
all_eliminated = false;
}
}
// Move this block to the final list if it has not been previously marked eliminated
if(!terminalIter->isEliminated)
{
#ifdef DEBUGOVERLAP
std::cout << "[II] Adding block " << *terminalIter << " to final list\n";
//std::cout << " extension: " << terminalIter->forwardHistory << "\n";
#endif
pOBFinal->push_back(*terminalIter);
}
}
}
activeList.clear();
}
OverlapResult OverlapAlgorithm::overlapReadInexact(const SeqRecord& read, int minOverlap, OverlapBlockList* pOBOut) const
{
OverlapResult result;
OverlapBlockList obWorkingList;
std::string seq = read.seq.toString();
#ifdef DEBUGOVERLAP
std::cout << "\n\n***Overlapping read " << read.id << " suffix\n";
#endif
// Match the suffix of seq to prefixes
// findInexact returns false is the maximum search time was exceeded. In this
// case we dont run any of the subsequent commands and return no overlaps.
bool valid = true;
valid = findOverlapBlocksInexact(seq, m_pBWT, m_pRevBWT, sufPreAF,
minOverlap, &obWorkingList, pOBOut, result);
if(valid)
valid = findOverlapBlocksInexact(complement(seq), m_pRevBWT, m_pBWT, prePreAF,
minOverlap, &obWorkingList, pOBOut, result);
if(valid)
{
if(m_bIrreducible)
{
computeIrreducibleBlocks(m_pBWT, m_pRevBWT, &obWorkingList, pOBOut);
obWorkingList.clear();
}
else
{
pOBOut->splice(pOBOut->end(), obWorkingList);
assert(obWorkingList.empty());
}
}
#ifdef DEBUGOVERLAP
std::cout << "\n\n***Overlapping read " << read.id << " prefix\n";
#endif
// Match the prefix of seq to suffixes
if(valid)
valid = findOverlapBlocksInexact(reverseComplement(seq), m_pBWT, m_pRevBWT, sufSufAF, minOverlap, &obWorkingList, pOBOut, result);
if(valid)
valid = findOverlapBlocksInexact(reverse(seq), m_pRevBWT, m_pBWT, preSufAF, minOverlap, &obWorkingList, pOBOut, result);
if(valid)
{
if(m_bIrreducible)
{
computeIrreducibleBlocks(m_pBWT, m_pRevBWT, &obWorkingList, pOBOut);
obWorkingList.clear();
}
else
{
pOBOut->splice(pOBOut->end(), obWorkingList);
assert(obWorkingList.empty());
}
}
if(!valid)
{
pOBOut->clear();
result.isSubstring = false;
result.searchAborted = true;
return result;
}
return result;
}