// Find the interval pair corresponding to w using a cached intervals for short substrings
BWTIntervalPair BWTAlgorithms::findIntervalPairWithCache(const BWT* pBWT,
const BWT* pRevBWT,
const BWTIntervalCache* pFwdCache,
const BWTIntervalCache* pRevCache,
const std::string& w)
{
size_t cacheLen = pFwdCache->getCachedLength();
if(w.size() < cacheLen)
return findIntervalPair(pBWT, pRevBWT, w);
// Compute the fwd and reverse interval using the cache for the last k bases
BWTIntervalPair ip;
int len = w.size();
int j = len - cacheLen;
std::string ss = w.substr(j);
std::string r_ss = reverse(ss);
assert(ss.size() == cacheLen);
ip.interval[0] = pFwdCache->lookup(ss.c_str());
ip.interval[1] = pRevCache->lookup(r_ss.c_str());
// Extend the interval to the full length of w as normal
j -= 1;
for(;j >= 0; --j)
{
updateBothL(ip, w[j], pBWT);
if(!ip.isValid())
return ip;
}
return ip;
}
// Find the intervals in pBWT/pRevBWT corresponding to w
// If w does not exist in the BWT, the interval
// coordinates [l, u] will be such that l > u
BWTIntervalPair BWTAlgorithms::findIntervalPair(const BWT* pBWT, const BWT* pRevBWT, const std::string& w)
{
BWTIntervalPair intervals;
int len = w.size();
int j = len - 1;
char curr = w[j];
initIntervalPair(intervals, curr, pBWT, pRevBWT);
--j;
for(;j >= 0; --j)
{
curr = w[j];
updateBothL(intervals, curr, pBWT);
if(!intervals.isValid())
return intervals;
}
return intervals;
}
// Calculate the 1-base de Bruijn graph extensions of str
// The includes the reverse complement
AlphaCount64 BWTAlgorithms::calculateDeBruijnExtensions(const std::string str,
const BWT* pBWT,
const BWT* pRevBWT,
EdgeDir direction,
const BWTIntervalCache* pFwdCache,
const BWTIntervalCache* pRevCache)
{
size_t k = str.size();
size_t p = k - 1;
std::string pmer;
// In the sense direction, we extend from the 3' end
if(direction == ED_SENSE)
pmer = str.substr(1, p);
else
pmer = str.substr(0, p);
assert(pmer.length() == p);
std::string rc_pmer = reverseComplement(pmer);
// Get the interval for the p-mer and its reverse complement
BWTIntervalPair ip;
BWTIntervalPair rc_ip;
// If pointers to interval caches are available, use them
// to speed up the initial calculation
if(pFwdCache != NULL && pRevCache != NULL)
{
ip = BWTAlgorithms::findIntervalPairWithCache(pBWT, pRevBWT, pFwdCache, pRevCache, pmer);
rc_ip = BWTAlgorithms::findIntervalPairWithCache(pBWT, pRevBWT, pFwdCache, pRevCache, rc_pmer);
}
else
{
ip = BWTAlgorithms::findIntervalPair(pBWT, pRevBWT, pmer);
rc_ip = BWTAlgorithms::findIntervalPair(pBWT, pRevBWT, rc_pmer);
}
assert(ip.isValid() || rc_ip.isValid());
// Get the extension bases
AlphaCount64 extensions;
AlphaCount64 rc_extensions;
// Calculate the interval to use to find the extensions. If extending in the sense
// direction this is the reverse interval/reverse bwt for the forward bwt and the forward
// interval for the reverse BWT. Vice-versa for anti-sense
size_t fwdIdx;
if(direction == ED_SENSE)
fwdIdx = 1;
else
fwdIdx = 0;
size_t revIdx = 1 - fwdIdx;
const BWT* bwts[2];
bwts[0] = pBWT;
bwts[1] = pRevBWT;
if(ip.interval[fwdIdx].isValid())
extensions += BWTAlgorithms::getExtCount(ip.interval[fwdIdx], bwts[fwdIdx]);
if(rc_ip.interval[revIdx].isValid())
rc_extensions = BWTAlgorithms::getExtCount(rc_ip.interval[revIdx], bwts[revIdx]);
// Switch the reverse-complement extensions to the same strand as the str
rc_extensions.complement();
extensions += rc_extensions;
return extensions;
}
// Calculate the ranges in pBWT that contain a prefix of at least minOverlap basepairs that
// overlaps with a suffix of w. The ranges are added to the pOBList
void OverlapAlgorithm::findOverlapBlocksExact(const std::string& w, const BWT* pBWT,
const BWT* pRevBWT, const AlignFlags& af, int minOverlap,
OverlapBlockList* pOverlapList, OverlapBlockList* pContainList,
OverlapResult& result) const
{
// The algorithm is as follows:
// We perform a backwards search using the FM-index for the string w.
// As we perform the search we collect the intervals
// of the significant prefixes (len >= minOverlap) that overlap w.
BWTIntervalPair ranges;
size_t l = w.length();
int start = l - 1;
BWTAlgorithms::initIntervalPair(ranges, w[start], pBWT, pRevBWT);
// Collect the OverlapBlocks
for(size_t i = start - 1; i >= 1; --i)
{
// Compute the range of the suffix w[i, l]
BWTAlgorithms::updateBothL(ranges, w[i], pBWT);
int overlapLen = l - i;
if(overlapLen >= minOverlap)
{
// Calculate which of the prefixes that match w[i, l] are terminal
// These are the proper prefixes (they are the start of a read)
BWTIntervalPair probe = ranges;
BWTAlgorithms::updateBothL(probe, '$', pBWT);
// The probe interval contains the range of proper prefixes
if(probe.interval[1].isValid())
{
assert(probe.interval[1].lower > 0);
pOverlapList->push_back(OverlapBlock(probe, ranges, overlapLen, 0, af));
}
}
}
// Determine if this sequence is contained and should not be processed further
BWTAlgorithms::updateBothL(ranges, w[0], pBWT);
// Ranges now holds the interval for the full-length read
// To handle containments, we output the overlapBlock to the final overlap block list
// and it will be processed later
// Two possible containment cases:
// 1) This read is a substring of some other read
// 2) This read is identical to some other read
// Case 1 is indicated by the existance of a non-$ left or right hand extension
// In this case we return no alignments for the string
AlphaCount64 left_ext = BWTAlgorithms::getExtCount(ranges.interval[0], pBWT);
AlphaCount64 right_ext = BWTAlgorithms::getExtCount(ranges.interval[1], pRevBWT);
if(left_ext.hasDNAChar() || right_ext.hasDNAChar())
{
result.isSubstring = true;
}
else
{
BWTIntervalPair probe = ranges;
BWTAlgorithms::updateBothL(probe, '$', pBWT);
if(probe.isValid())
{
// terminate the contained block and add it to the contained list
BWTAlgorithms::updateBothR(probe, '$', pRevBWT);
assert(probe.isValid());
pContainList->push_back(OverlapBlock(probe, ranges, w.length(), 0, af));
}
}
//OverlapBlockList containedWorkingList;
//partitionBlockList(w.length(), &workingList, pOverlapList, &containedWorkingList);
// Terminate the contained blocks
//terminateContainedBlocks(containedWorkingList);
// Move the contained blocks to the final contained list
//pContainList->splice(pContainList->end(), containedWorkingList);
return;
}
