// Extend all the blocks in activeList by one base to the right
// Move all right-terminal blocks to the termainl list. If a block
// is terminal and potentially contained by another block, add it to
// containedList
void OverlapAlgorithm::extendActiveBlocksRight(const BWT* pBWT, const BWT* pRevBWT,
OverlapBlockList& activeList,
OverlapBlockList& terminalList,
OverlapBlockList& /*containedList*/) const
{
OverlapBlockList::iterator iter = activeList.begin();
OverlapBlockList::iterator next;
while(iter != activeList.end())
{
next = iter;
++next;
// Check if block is terminal
AlphaCount64 ext_count = iter->getCanonicalExtCount(pBWT, pRevBWT);
if(ext_count.get('$') > 0)
{
// Only consider this block to be terminal irreducible if it has at least one extension
// or else it is a substring block
if(iter->forwardHistory.size() > 0)
{
OverlapBlock branched = *iter;
BWTAlgorithms::updateBothR(branched.ranges, '$', branched.getExtensionBWT(pBWT, pRevBWT));
terminalList.push_back(branched);
#ifdef DEBUGOVERLAP_2
std::cout << "Block of length " << iter->overlapLen << " moved to terminal\n";
#endif
}
}
int curr_extension = iter->forwardHistory.size();
// Perform the right extensions
// Best case, there is only a single extension character
// Handle this case specially so we don't need to copy the potentially
// large OverlapBlock structure and its full history
if(ext_count.hasUniqueDNAChar())
{
// Get the extension character with respect to the queried sequence
char canonical_base = ext_count.getUniqueDNAChar();
// Flip the base into the frame of reference for the block
char block_base = iter->flags.isQueryComp() ? complement(canonical_base) : canonical_base;
// Update the block using the base in its frame of reference
BWTAlgorithms::updateBothR(iter->ranges, block_base, iter->getExtensionBWT(pBWT, pRevBWT));
// Add the base to the history in the frame of reference of the query read
// This is so the history is consistent when comparing between blocks from different strands
iter->forwardHistory.add(curr_extension, canonical_base);
}
else
{
for(size_t idx = 0; idx < DNA_ALPHABET_SIZE; ++idx)
{
char canonical_base = ALPHABET[idx];
char block_base = iter->flags.isQueryComp() ? complement(canonical_base) : canonical_base;
if(ext_count.get(canonical_base) == 0)
continue;
// Branch the sequence. This involves copying the entire history which can be large
// if the input sequences are very long. This could be avoided by using the SearchHistoyNode/Link
// structure but branches are infrequent enough to not have a large impact
OverlapBlock branched = *iter;
BWTAlgorithms::updateBothR(branched.ranges, block_base, branched.getExtensionBWT(pBWT, pRevBWT));
assert(branched.ranges.isValid());
// Add the base in the canonical frame
branched.forwardHistory.add(curr_extension, canonical_base);
// Insert the new block after the iterator
activeList.insert(iter, branched);
}
// Remove the original block, which has been superceded by the branches
activeList.erase(iter);
}
iter = next; // this skips the newly-inserted blocks
}
}
// Construct the set of blocks describing irreducible overlaps with READ
// and write the blocks to pOBOut
OverlapResult OverlapAlgorithm::overlapReadExact(const SeqRecord& read, int minOverlap, OverlapBlockList* pOBOut) const
{
OverlapResult result;
// The complete set of overlap blocks are collected in obWorkingList
// The filtered set (containing only irreducible overlaps) are placed into pOBOut
// by calculateIrreducibleHits
OverlapBlockList obWorkingList;
std::string seq = read.seq.toString();
// We store the various overlap blocks using a number of lists, one for the containments
// in the forward and reverse index and one for each set of overlap blocks
OverlapBlockList oblFwdContain;
OverlapBlockList oblRevContain;
OverlapBlockList oblSuffixFwd;
OverlapBlockList oblSuffixRev;
OverlapBlockList oblPrefixFwd;
OverlapBlockList oblPrefixRev;
// Match the suffix of seq to prefixes
findOverlapBlocksExact(seq, m_pBWT, m_pRevBWT, sufPreAF, minOverlap, &oblSuffixFwd, &oblFwdContain, result);
if (!m_noReverse)
{
findOverlapBlocksExact(complement(seq), m_pRevBWT, m_pBWT, prePreAF, minOverlap, &oblSuffixRev, &oblRevContain, result);
}
// Match the prefix of seq to suffixes
if (!m_noReverse)
{
findOverlapBlocksExact(reverseComplement(seq), m_pBWT, m_pRevBWT, sufSufAF, minOverlap, &oblPrefixFwd, &oblFwdContain, result);
}
findOverlapBlocksExact(reverse(seq), m_pRevBWT, m_pBWT, preSufAF, minOverlap, &oblPrefixRev, &oblRevContain, result);
// Remove submaximal blocks for each block list including fully contained blocks
// Copy the containment blocks into the prefix/suffix lists
oblSuffixFwd.insert(oblSuffixFwd.end(), oblFwdContain.begin(), oblFwdContain.end());
oblPrefixFwd.insert(oblPrefixFwd.end(), oblFwdContain.begin(), oblFwdContain.end());
oblSuffixRev.insert(oblSuffixRev.end(), oblRevContain.begin(), oblRevContain.end());
oblPrefixRev.insert(oblPrefixRev.end(), oblRevContain.begin(), oblRevContain.end());
// Perform the submaximal filter
removeSubMaximalBlocks(&oblSuffixFwd, m_pBWT, m_pRevBWT);
removeSubMaximalBlocks(&oblPrefixFwd, m_pBWT, m_pRevBWT);
removeSubMaximalBlocks(&oblSuffixRev, m_pRevBWT, m_pBWT);
removeSubMaximalBlocks(&oblPrefixRev, m_pRevBWT, m_pBWT);
// Remove the contain blocks from the suffix/prefix lists
removeContainmentBlocks(seq.length(), &oblSuffixFwd);
removeContainmentBlocks(seq.length(), &oblPrefixFwd);
removeContainmentBlocks(seq.length(), &oblSuffixRev);
removeContainmentBlocks(seq.length(), &oblPrefixRev);
// Join the suffix and prefix lists
oblSuffixFwd.splice(oblSuffixFwd.end(), oblSuffixRev);
oblPrefixFwd.splice(oblPrefixFwd.end(), oblPrefixRev);
// Move the containments to the output list
pOBOut->splice(pOBOut->end(), oblFwdContain);
pOBOut->splice(pOBOut->end(), oblRevContain);
// Filter out transitive overlap blocks if requested
if(m_bIrreducible)
{
computeIrreducibleBlocks(m_pBWT, m_pRevBWT, &oblSuffixFwd, pOBOut);
computeIrreducibleBlocks(m_pBWT, m_pRevBWT, &oblPrefixFwd, pOBOut);
}
else
{
pOBOut->splice(pOBOut->end(), oblSuffixFwd);
pOBOut->splice(pOBOut->end(), oblPrefixFwd);
}
return result;
}