int LocateAnchorBoundsInSuffixArray(T_RefSequence &reference,
T_SuffixArray &sa, T_Sequence &read, unsigned int minPrefixMatchLength,
std::vector<DNALength> &matchLow, std::vector<DNALength> &matchHigh,
std::vector<DNALength> &matchLength, AnchorParameters ¶ms) {
//
// Make sure there is enough of this read to map. Since searches
// are keyed off of 'minPrefixMatchLength' matches, don't search
// anything shorter than that.
//
if (minPrefixMatchLength > 0 and
read.SubreadLength() < minPrefixMatchLength) {
return 0;
}
DNALength p, m;
DNALength matchEnd = read.SubreadEnd() - minPrefixMatchLength + 1;
DNALength numSearchedPositions = matchEnd - read.SubreadStart();
matchLength.resize(numSearchedPositions);
matchLow.resize(numSearchedPositions);
matchHigh.resize(numSearchedPositions);
std::fill(matchLength.begin(), matchLength.end(), 0);
std::fill(matchLow.begin(), matchLow.end(), 0);
std::fill(matchHigh.begin(), matchHigh.end(), 0);
vector<SAIndex> lowMatchBound, highMatchBound;
for (m = 0, p = read.SubreadStart(); p < matchEnd; p++, m++) {
lowMatchBound.clear(); highMatchBound.clear();
DNALength lcpLength = sa.StoreLCPBounds(reference.seq, reference.length,
&read.seq[p], matchEnd - p,
params.useLookupTable,
params.maxLCPLength,
//
// Store the positions in the SA
// that are searched.
//
lowMatchBound, highMatchBound,
params.stopMappingOnceUnique);
//
// Possibly print the lcp bounds for debugging
//
if (params.lcpBoundsOutPtr != NULL) {
for (size_t i = 0; i < lowMatchBound.size(); i++) {
*params.lcpBoundsOutPtr <<
(highMatchBound[i] - lowMatchBound[i]);
if (i < lowMatchBound.size() - 1) {
*params.lcpBoundsOutPtr << " ";
}
}
*params.lcpBoundsOutPtr << endl;
}
//
// Default to no match.
//
matchLow[m] = matchHigh[m] = matchLength[m] = 0;
//
// If anything was found in the suffix array:
//
if (lowMatchBound.size() > 0) {
//
// First expand the search bounds until at least
// one match is found.
//
int lcpSearchLength = lowMatchBound.size();
while (lcpSearchLength > 0 and
lowMatchBound[lcpSearchLength - 1] ==
highMatchBound[lcpSearchLength - 1]) {
lcpSearchLength--;
lcpLength--;
}
matchLow[m] = lowMatchBound[lcpSearchLength - 1];
matchHigh[m] = highMatchBound[lcpSearchLength - 1];
matchLength[m] = minPrefixMatchLength + lcpSearchLength;
//
// Next, apply some heuristics to the anchor generation.
//
// 1.1 If the suffix array match is unique, try and extend that
// match as long as possible to ease global chaining later on.
//
// 1.2 If the suffix array match is unique, but cannot be
// extended, it probably ends in an error. Back the search up
// by 1.
//
// 2.1 If the suffix array match is not unique, return the
// default matches, or expand the search to include more
// matches.
//
//
// Check to see if the match was unique.
//
if (matchLow[m] + 1 == matchHigh[m]) {
//
// If the match is unique, extend for as long as possible.
//
lcpLength = minPrefixMatchLength + lcpSearchLength;
long refPos = sa.index[matchLow[m]] + lcpLength;
long queryPos = p + lcpLength;
bool extensionWasPossible = false;
while (refPos + 1 < reference.length and
queryPos + 1 < read.length and
reference.seq[refPos + 1] == read.seq[queryPos + 1] and
(params.maxLCPLength == 0 or
lcpLength < static_cast<DNALength>(params.maxLCPLength))) {
refPos++;
queryPos++;
lcpLength++;
extensionWasPossible = true;
}
if (extensionWasPossible) {
//
// Was able to extend match far into the genome, store that.
//
matchLength[m] = lcpLength;
}
else if (extensionWasPossible == false) {
//
// No extension was possible, indicating that this match
// ends at an error. To be safe, expand search by up to
// 1.
//
if (lcpSearchLength > 1) {
lcpSearchLength = lcpSearchLength - 1;
}
matchLow[m] = lowMatchBound[lcpSearchLength-1];
matchHigh[m] = highMatchBound[lcpSearchLength-1];
matchLength[m] = minPrefixMatchLength + lcpSearchLength;
}
}
else {
//
// The match is not unique. Store a possibly expanded search.
//
if (lcpSearchLength > params.expand) {
lcpSearchLength -= params.expand;
}
else {
assert(lowMatchBound.size() > 0);
lcpSearchLength = 1;
}
//
// There are multiple matches for this position.
//
matchLow[m] = lowMatchBound[lcpSearchLength - 1];
matchHigh[m] = highMatchBound[lcpSearchLength - 1];
matchLength[m] = minPrefixMatchLength + lcpSearchLength;
}
}
else {
//
// The match is shorter than what the search is supposed to
// expand to. In order to avoid expanding to before the end
// of the match list, do not set any match.
//
matchLow[m] = 0;
matchHigh[m] = 0;
matchLength[m] = 0;
}
//
// Possibly advance a bunch of steps.
//
if (params.advanceExactMatches) {
int tmp = (int)lcpLength - (int)params.expand
- params.advanceExactMatches;
int advance = MAX(tmp, 0);
p += advance;
m += advance;
}
}
return 1;
}
void SAMOutput::CreateCIGARString(T_AlignmentCandidate &alignment,
T_Sequence &read,
std::string &cigarString,
Clipping clipping,
DNALength & prefixSoftClip, DNALength & suffixSoftClip,
DNALength & prefixHardClip, DNALength & suffixHardClip,
bool cigarUseSeqMatch, const bool allowAdjacentIndels) {
cigarString = "";
// All cigarString use the no clipping core
std::vector<int> opSize;
std::vector<char> opChar;
CreateNoClippingCigarOps(alignment, opSize, opChar, cigarUseSeqMatch, allowAdjacentIndels);
// Clipping needs to be added
if (clipping == hard) {
SetHardClip(alignment, read, prefixHardClip, suffixHardClip);
if (prefixHardClip > 0) {
opSize.insert(opSize.begin(), prefixHardClip);
opChar.insert(opChar.begin(), 'H');
}
if (suffixHardClip > 0) {
opSize.push_back(suffixHardClip);
opChar.push_back('H');
}
prefixSoftClip = 0;
suffixSoftClip = 0;
}
if (clipping == soft or clipping == subread) {
//
// Even if clipping is soft, the hard clipping removes the
// low quality regions
//
if (clipping == soft) {
prefixHardClip = read.lowQualityPrefix;
suffixHardClip = read.lowQualitySuffix;
}
else if (clipping == subread) {
prefixHardClip = std::max((DNALength) read.SubreadStart(), read.lowQualityPrefix);
suffixHardClip = std::max((DNALength)(read.length - read.SubreadEnd()), read.lowQualitySuffix);
}
SetSoftClip(alignment, read, prefixHardClip, suffixHardClip, prefixSoftClip, suffixSoftClip);
if (alignment.tStrand == 1) {
std::swap(prefixHardClip, suffixHardClip);
std::swap(prefixSoftClip, suffixSoftClip);
}
//
// Insert the hard and soft clipping so that they are in the
// order H then S if both exist.
//
if (prefixSoftClip > 0) {
opSize.insert(opSize.begin(), prefixSoftClip);
opChar.insert(opChar.begin(), 'S');
}
if (prefixHardClip > 0) {
opSize.insert(opSize.begin(), prefixHardClip);
opChar.insert(opChar.begin(), 'H');
}
//
// Append the hard and soft clipping so they are in the order S
// then H.
//
if (suffixSoftClip > 0) {
opSize.push_back(suffixSoftClip);
opChar.push_back('S');
}
if (suffixHardClip > 0) {
opSize.push_back(suffixHardClip);
opChar.push_back('H');
}
}
CigarOpsToString(opSize, opChar, cigarString);
}
int MapReadToGenome(T_RefSequence &reference,
T_SuffixArray &sa, T_Sequence &read,
unsigned int minPrefixMatchLength,
vector<T_MatchPos> &matchPosList,
AnchorParameters &anchorParameters) {
vector<DNALength> matchLow, matchHigh, matchLength;
DNALength minMatchLen = anchorParameters.minMatchLength;
if (read.SubreadLength() < minMatchLen) {
matchPosList.clear();
return 0;
}
LocateAnchorBoundsInSuffixArray(reference, sa, read,
minPrefixMatchLength, matchLow, matchHigh, matchLength,
anchorParameters);
//
// Try evaluating some contexts.
//
DNALength pos;
assert(matchLow.size() == matchHigh.size());
DNASequence evalQrySeq, evalRefSeq;
vector<Arrow> pathMat;
vector<int> scoreMat;
Alignment alignment;
//
// Do some filtering on the matches looking for overlapping matches
// if there are any.
//
if (anchorParameters.removeEncompassedMatches) {
vector<bool> removed;
removed.resize(read.length);
std::fill(removed.begin(), removed.end(), false);
size_t i;
for (i = 0; i < read.length-1; i++) {
if (matchLength[i] == matchLength[i+1]+1) {
removed[i+1] = true;
}
}
for (i = 1; i < matchLength.size(); i++) {
if (removed[i]) {
matchLength[i] = matchLow[i] = matchHigh[i] = 0;
}
}
}
//
// Now add
//
DNALength endOfMapping;
DNALength trim = MAX(minMatchLen + 1, sa.lookupPrefixLength + 1);
if (read.SubreadEnd() < trim) {
endOfMapping = 0;
}
else {
endOfMapping = read.SubreadEnd() - trim;
}
for (pos = read.SubreadStart(); pos < endOfMapping; pos++) {
size_t matchIndex = pos - read.SubreadStart();
assert(matchIndex < matchHigh.size());
if (matchHigh[matchIndex] - matchLow[matchIndex] <=
anchorParameters.maxAnchorsPerPosition) {
DNALength mp;
for (mp = matchLow[matchIndex]; mp < matchHigh[matchIndex]; mp++) {
if (matchLength[matchIndex] < minMatchLen) {
continue;
}
//
// By default, add all anchors.
//
if (matchLength[matchIndex] + pos > read.length) {
//
// When doing branching, it's possible that a deletion
// branch finds an anchor that goes past the end of a
// read. When that is the case, trim back the anchor
// match since this confuses downstream assertions.
//
matchLength[matchIndex] = read.length - pos;
}
assert(sa.index[mp] + matchLength[matchIndex]
<= reference.length);
matchPosList.push_back(ChainedMatchPos(sa.index[mp], pos,
matchLength[matchIndex],
matchHigh[matchIndex] - matchLow[matchIndex]));
}
}
}
return matchPosList.size();
}
void SAMOutput::PrintAlignment(T_AlignmentCandidate &alignment,
T_Sequence &read,
std::ostream &samFile,
AlignmentContext &context,
SupplementalQVList & qvList,
Clipping clipping,
bool cigarUseSeqMatch,
const bool allowAdjacentIndels) {
std::string cigarString;
uint16_t flag;
T_Sequence alignedSequence;
DNALength prefixSoftClip = 0, suffixSoftClip = 0;
DNALength prefixHardClip = 0, suffixHardClip = 0;
CreateCIGARString(alignment, read, cigarString, clipping, prefixSoftClip, suffixSoftClip, prefixHardClip, suffixHardClip, cigarUseSeqMatch, allowAdjacentIndels);
SetAlignedSequence(alignment, read, alignedSequence, clipping);
BuildFlag(alignment, context, flag);
samFile << alignment.qName << "\t"
<< flag << "\t"
<< alignment.tName << "\t"; // RNAME
if (alignment.tStrand == 0) {
samFile << alignment.TAlignStart() + 1 << "\t";
// POS, add 1 to get 1 based coordinate system
}
else {
samFile << alignment.tLength - (alignment.TAlignStart() + alignment.TEnd()) + 1 << "\t"; // includes - 1 for rev-comp, +1 for one-based
}
samFile << (int) alignment.mapQV << "\t"// MAPQ
<< cigarString << "\t"; // CIGAR
//
// Determine RNEXT
std::string rNext;
rNext = "*";
/*
if (context.hasNextSubreadPos == false) {
rNext = "*";
}
else {
if (context.rNext == alignment.tName) {
rNext = "=";
}
else {
rNext = context.rNext;
}
}
*/
samFile << rNext << "\t"; // RNEXT
DNALength nextSubreadPos = 0;
/*
if (context.hasNextSubreadPos) {
nextSubreadPos = context.nextSubreadPos + 1;
}*/
samFile << nextSubreadPos << "\t"; // RNEXT, add 1 for 1 based
// indexing
//DNALength tLen = alignment.GenomicTEnd() - alignment.GenomicTBegin();
//SAM v1.5, tLen is set as 0 for single-segment template
samFile << 0 << "\t"; // TLEN
// Print the sequence on one line, and suppress printing the
// newline (by setting the line length to alignedSequence.length
(static_cast<DNASequence*>(&alignedSequence))->PrintSeq(samFile, 0); // SEQ
samFile << "\t";
if (alignedSequence.qual.data != NULL && qvList.useqv == 0) {
alignedSequence.PrintAsciiQual(samFile, 0); // QUAL
}
else {
samFile <<"*";
}
samFile << "\t";
//
// Add optional fields
//
samFile << "RG:Z:" << context.readGroupId << "\t";
samFile << "AS:i:" << alignment.score << "\t";
//
// "RG" read group Id
// "AS" alignment score
// "XS" read alignment start position without counting previous soft clips (1 based)
// "XE" read alignment end position without counting previous soft clips (1 based)
// "XL" aligned read length
// "XQ" query sequence length
// "XT" # of continues reads, always 1 for blasr
// "NM" edit distance
// "FI" read alignment start position (1 based)
//
DNALength qAlignStart = alignment.QAlignStart();
DNALength qAlignEnd = alignment.QAlignEnd();
if (clipping == none) {
samFile << "XS:i:" << qAlignStart + 1 << "\t";
samFile << "XE:i:" << qAlignEnd + 1 << "\t";
}
else if (clipping == hard or clipping == soft or clipping == subread) {
DNALength xs = prefixHardClip;
DNALength xe = read.length - suffixHardClip;
if (alignment.tStrand == 1) {
xs = suffixHardClip;
xe = read.length - prefixHardClip;
}
samFile << "XS:i:" << xs + 1 << "\t"; // add 1 for 1-based indexing in sam
assert(read.length - suffixHardClip == prefixHardClip + alignedSequence.length);
samFile << "XE:i:" << xe + 1 << "\t";
}
samFile << "YS:i:" << read.SubreadStart() << "\t";
samFile << "YE:i:" << read.SubreadEnd() << "\t";
samFile << "ZM:i:" << read.HoleNumber() << "\t";
samFile << "XL:i:" << alignment.qAlignedSeq.length << "\t";
samFile << "XT:i:1\t"; // reads are allways continuous reads, not
// referenced based circular consensus when
// output by blasr.
samFile << "NM:i:" << context.editDist << "\t";
samFile << "FI:i:" << alignment.qAlignedSeqPos + 1;
// Add query sequence length
samFile << "\t" << "XQ:i:" << alignment.qLength;
//
// Write out optional quality values. If qvlist does not
// have any qv's signaled to print, this is a no-op.
//
// First transform characters that are too large to printable ones.
qvList.FormatQVOptionalFields(alignedSequence);
qvList.PrintQVOptionalFields(alignedSequence, samFile);
samFile << std::endl;
}