TEST(CigarStringTest, FromStdString_MultipleOps)
{
const string multiCigar = "100=2D34I6=6X6=";
Cigar cigar = Cigar::FromStdString(multiCigar);
ASSERT_TRUE(cigar.size() == 6);
CigarOperation op0 = cigar.at(0);
CigarOperation op1 = cigar.at(1);
CigarOperation op2 = cigar.at(2);
CigarOperation op3 = cigar.at(3);
CigarOperation op4 = cigar.at(4);
CigarOperation op5 = cigar.at(5);
EXPECT_TRUE(op0.Char() == '=');
EXPECT_TRUE(op0.Length() == 100);
EXPECT_TRUE(op1.Char() == 'D');
EXPECT_TRUE(op1.Length() == 2);
EXPECT_TRUE(op2.Char() == 'I');
EXPECT_TRUE(op2.Length() == 34);
EXPECT_TRUE(op3.Char() == '=');
EXPECT_TRUE(op3.Length() == 6);
EXPECT_TRUE(op4.Char() == 'X');
EXPECT_TRUE(op4.Length() == 6);
EXPECT_TRUE(op5.Char() == '=');
EXPECT_TRUE(op5.Length() == 6);
}
unsigned UngappedAligner::alignUngapped(
FragmentMetadata &fragmentMetadata,
Cigar &cigarBuffer,
const flowcell::ReadMetadata &readMetadata,
const matchSelector::FragmentSequencingAdapterClipper &adapterClipper,
const reference::ContigList &contigList,
const isaac::reference::ContigAnnotations &contigAnnotations) const
{
const unsigned cigarOffset = cigarBuffer.size();
// Don't reset alignment to preserve the seed-based anchors.
// fragmentMetadata.resetAlignment();
ISAAC_ASSERT_MSG(!fragmentMetadata.isAligned(), "alignUngapped is expected to be performend on a clean fragment");
fragmentMetadata.resetClipping();
const reference::Contig &contig = contigList[fragmentMetadata.contigId];
const Read &read = fragmentMetadata.getRead();
const bool reverse = fragmentMetadata.reverse;
const std::vector<char> &sequence = read.getStrandSequence(reverse);
const reference::Contig &reference = contig;
std::vector<char>::const_iterator sequenceBegin = sequence.begin();
std::vector<char>::const_iterator sequenceEnd = sequence.end();
adapterClipper.clip(contig, fragmentMetadata, sequenceBegin, sequenceEnd);
clipReadMasking(read, fragmentMetadata, sequenceBegin, sequenceEnd);
clipReference(reference.size(), fragmentMetadata, sequenceBegin, sequenceEnd);
const unsigned firstMappedBaseOffset = std::distance(sequence.begin(), sequenceBegin);
if (firstMappedBaseOffset)
{
cigarBuffer.addOperation(firstMappedBaseOffset, Cigar::SOFT_CLIP);
}
const unsigned mappedBases = std::distance(sequenceBegin, sequenceEnd);
if (mappedBases)
{
const Cigar::OpCode opCode = Cigar::ALIGN;
cigarBuffer.addOperation(mappedBases, opCode);
}
const unsigned clipEndBases = std::distance(sequenceEnd, sequence.end());
if (clipEndBases)
{
cigarBuffer.addOperation(clipEndBases, Cigar::SOFT_CLIP);
}
const unsigned ret = updateFragmentCigar(
readMetadata, contigList, contigAnnotations, fragmentMetadata,
fragmentMetadata.reverse, fragmentMetadata.contigId, fragmentMetadata.position, cigarBuffer, cigarOffset);
if (!ret)
{
fragmentMetadata.setUnaligned();
}
return ret;
}
TEST(CigarStringTest, ToStdString_SingleOp)
{
const string singleCigar = "100=";
Cigar cigar;
cigar.push_back( CigarOperation(CigarOperationType::SEQUENCE_MATCH, 100) );
EXPECT_EQ(singleCigar, cigar.ToStdString());
}
int getMaxClipLen( SamRecord & sam_rec )
{
Cigar * myCigar = sam_rec.getCigarInfo();
int begin_clip = myCigar->getNumBeginClips();
int end_clip = myCigar->getNumEndClips();
if (begin_clip >= end_clip)
return begin_clip;
else
return -end_clip;
}
TEST(CigarStringTest, FromStdString_SingleOp)
{
const string singleCigar = "100=";
Cigar cigar = Cigar::FromStdString(singleCigar);
ASSERT_TRUE(cigar.size() == 1);
const CigarOperation& op = cigar.front();
EXPECT_TRUE(op.Char() == '=');
EXPECT_TRUE(op.Length() == 100);
}
TEST(CigarStringTest, ToStdString_MultipleOps)
{
const string multiCigar = "100=2D34I6=6X6=";
Cigar cigar;
cigar.push_back(CigarOperation(CigarOperationType::SEQUENCE_MATCH, 100));
cigar.push_back(CigarOperation(CigarOperationType::DELETION, 2));
cigar.push_back(CigarOperation(CigarOperationType::INSERTION, 34));
cigar.push_back(CigarOperation(CigarOperationType::SEQUENCE_MATCH, 6));
cigar.push_back(CigarOperation(CigarOperationType::SEQUENCE_MISMATCH, 6));
cigar.push_back(CigarOperation(CigarOperationType::SEQUENCE_MATCH, 6));
EXPECT_EQ(multiCigar, cigar.ToStdString());
}
void Sites::addToCurrentCluster( vector<bool> & is_in_coord, SingleSite & new_site, SamRecord & rec )
{
if (is_in_coord.size() != NMEI)
morphError("[Sites::setNewCluster] is_in_coord size error");
// update breakpoint
int old_evi = new_site.evidence;
float a1 = (float)1 / float(old_evi+1);
int ep = getEstimatedBreakPoint(rec);
new_site.breakp = round( a1 * (float)ep + (float)new_site.breakp * (1-a1));
new_site.evidence++;
// update position
if (rec.get1BasedPosition() < new_site.start)
new_site.start = rec.get1BasedPosition();
else if (rec.get1BasedAlignmentEnd() > new_site.end)
new_site.end = rec.get1BasedAlignmentEnd();
// update info
if (rec.getFlag() & 0x10) {
if (new_site.right_clip_only) {
Cigar * myCigar = rec.getCigarInfo();
int begin_clip = myCigar->getNumBeginClips();
if ( begin_clip < MIN_CLIP/2)
new_site.right_clip_only = 0;
}
for(int m=0; m<NMEI; m++) {
if (is_in_coord[m])
new_site.right[m]++;
}
}
else {
if (new_site.left_clip_only) {
Cigar * myCigar = rec.getCigarInfo();
int end_clip = myCigar->getNumEndClips();
if (end_clip < MIN_CLIP/2)
new_site.left_clip_only = 0;
}
for( int m=0; m<NMEI; m++) {
if (is_in_coord[m])
new_site.left[m]++;
}
}
}
void Sites::setNewCluster( vector<bool> & is_in_coord, SingleSite & new_site, SamRecord & rec )
{
if (is_in_coord.size() != NMEI)
morphError("[Sites::setNewCluster] is_in_coord size error");
// set info
new_site.breakp = getEstimatedBreakPoint(rec);
new_site.rcount = 1;
new_site.evidence = 1;
for(int m=0; m<NMEI; m++) {
new_site.left[m] = 0;
new_site.right[m] = 0;
}
new_site.left_clip_only = 1;
new_site.right_clip_only = 1;
new_site.depth = current_depth;
new_site.depth_add = 1;
// set position & mtype
if ( rec.getFlag() & 0x10 ) { // right anchor
new_site.start = rec.get1BasedPosition();
new_site.end = rec.get1BasedAlignmentEnd();
Cigar * myCigar = rec.getCigarInfo();
int begin_clip = myCigar->getNumBeginClips();
if ( begin_clip < MIN_CLIP/2)
new_site.right_clip_only = 0;
for(int m=0; m<NMEI; m++) {
if (is_in_coord[m])
new_site.right[m] = 1;
}
}
else {
new_site.start = rec.get1BasedPosition();
new_site.end = rec.get1BasedAlignmentEnd();
Cigar * myCigar = rec.getCigarInfo();
int end_clip = myCigar->getNumEndClips();
if (end_clip < MIN_CLIP/2)
new_site.left_clip_only = 0;
for(int m=0; m<NMEI; m++) {
if (is_in_coord[m])
new_site.left[m] = 1;
}
}
}
bool leftAlign(string& alternateSequence, Cigar& cigar, string& referenceSequence, bool debug = false) {
int arsOffset = 0; // pointer to insertion point in aligned reference sequence
string alignedReferenceSequence = referenceSequence;
int aabOffset = 0;
string alignmentAlignedBases = alternateSequence;
// store information about the indels
vector<VCFIndelAllele> indels;
int rp = 0; // read position, 0-based relative to read
int sp = 0; // sequence position
string softBegin;
string softEnd;
stringstream cigar_before, cigar_after;
for (vector<pair<int, string> >::const_iterator c = cigar.begin();
c != cigar.end(); ++c) {
unsigned int l = c->first;
char t = c->second.at(0);
cigar_before << l << t;
if (t == 'M') { // match or mismatch
sp += l;
rp += l;
} else if (t == 'D') { // deletion
indels.push_back(VCFIndelAllele(false, l, sp, rp, referenceSequence.substr(sp, l)));
alignmentAlignedBases.insert(rp + aabOffset, string(l, '-'));
aabOffset += l;
sp += l; // update reference sequence position
} else if (t == 'I') { // insertion
indels.push_back(VCFIndelAllele(true, l, sp, rp, alternateSequence.substr(rp, l)));
alignedReferenceSequence.insert(sp + softBegin.size() + arsOffset, string(l, '-'));
arsOffset += l;
rp += l;
} else if (t == 'S') { // soft clip, clipped sequence present in the read not matching the reference
// remove these bases from the refseq and read seq, but don't modify the alignment sequence
if (rp == 0) {
alignedReferenceSequence = string(l, '*') + alignedReferenceSequence;
softBegin = alignmentAlignedBases.substr(0, l);
} else {
alignedReferenceSequence = alignedReferenceSequence + string(l, '*');
softEnd = alignmentAlignedBases.substr(alignmentAlignedBases.size() - l, l);
}
rp += l;
} else if (t == 'H') { // hard clip on the read, clipped sequence is not present in the read
} else if (t == 'N') { // skipped region in the reference not present in read, aka splice
sp += l;
}
}
int alignedLength = sp;
VCFLEFTALIGN_DEBUG("| " << cigar_before.str() << endl
<< "| " << alignedReferenceSequence << endl
<< "| " << alignmentAlignedBases << endl);
// if no indels, return the alignment
if (indels.empty()) { return false; }
// for each indel, from left to right
// while the indel sequence repeated to the left and we're not matched up with the left-previous indel
// move the indel left
vector<VCFIndelAllele>::iterator previous = indels.begin();
for (vector<VCFIndelAllele>::iterator id = indels.begin(); id != indels.end(); ++id) {
// left shift by repeats
//
// from 1 base to the length of the indel, attempt to shift left
// if the move would cause no change in alignment optimality (no
// introduction of mismatches, and by definition no change in gap
// length), move to the new position.
// in practice this moves the indel left when we reach the size of
// the repeat unit.
//
int steppos, readsteppos;
VCFIndelAllele& indel = *id;
int i = 1;
while (i <= indel.length) {
int steppos = indel.position - i;
int readsteppos = indel.readPosition - i;
#ifdef VERBOSE_DEBUG
if (debug) {
if (steppos >= 0 && readsteppos >= 0) {
cerr << referenceSequence.substr(steppos, indel.length) << endl;
cerr << alternateSequence.substr(readsteppos, indel.length) << endl;
cerr << indel.sequence << endl;
}
}
#endif
while (steppos >= 0 && readsteppos >= 0
&& indel.sequence == referenceSequence.substr(steppos, indel.length)
&& indel.sequence == alternateSequence.substr(readsteppos, indel.length)
&& (id == indels.begin()
|| (previous->insertion && steppos >= previous->position)
|| (!previous->insertion && steppos >= previous->position + previous->length))) {
VCFLEFTALIGN_DEBUG((indel.insertion ? "insertion " : "deletion ") << indel << " shifting " << i << "bp left" << endl);
indel.position -= i;
indel.readPosition -= i;
steppos = indel.position - i;
readsteppos = indel.readPosition - i;
}
do {
++i;
} while (i <= indel.length && indel.length % i != 0);
}
// left shift indels with exchangeable flanking sequence
//
// for example:
//
// GTTACGTT GTTACGTT
// GT-----T ----> G-----TT
//
// GTGTGACGTGT GTGTGACGTGT
// GTGTG-----T ----> GTG-----TGT
//
// GTGTG-----T GTG-----TGT
// GTGTGACGTGT ----> GTGTGACGTGT
//
//
steppos = indel.position - 1;
readsteppos = indel.readPosition - 1;
while (steppos >= 0 && readsteppos >= 0
&& alternateSequence.at(readsteppos) == referenceSequence.at(steppos)
&& alternateSequence.at(readsteppos) == indel.sequence.at(indel.sequence.size() - 1)
&& (id == indels.begin()
|| (previous->insertion && indel.position - 1 >= previous->position)
|| (!previous->insertion && indel.position - 1 >= previous->position + previous->length))) {
VCFLEFTALIGN_DEBUG((indel.insertion ? "insertion " : "deletion ") << indel << " exchanging bases " << 1 << "bp left" << endl);
indel.sequence = indel.sequence.at(indel.sequence.size() - 1) + indel.sequence.substr(0, indel.sequence.size() - 1);
indel.position -= 1;
indel.readPosition -= 1;
steppos = indel.position - 1;
readsteppos = indel.readPosition - 1;
}
// tracks previous indel, so we don't run into it with the next shift
previous = id;
}
// bring together floating indels
// from left to right
// check if we could merge with the next indel
// if so, adjust so that we will merge in the next step
if (indels.size() > 1) {
previous = indels.begin();
for (vector<VCFIndelAllele>::iterator id = (indels.begin() + 1); id != indels.end(); ++id) {
VCFIndelAllele& indel = *id;
// parsimony: could we shift right and merge with the previous indel?
// if so, do it
int prev_end_ref = previous->insertion ? previous->position : previous->position + previous->length;
int prev_end_read = !previous->insertion ? previous->readPosition : previous->readPosition + previous->length;
if (previous->insertion == indel.insertion
&& ((previous->insertion
&& (previous->position < indel.position
&& previous->readPosition + previous->readPosition < indel.readPosition))
||
(!previous->insertion
&& (previous->position + previous->length < indel.position)
&& (previous->readPosition < indel.readPosition)
))) {
if (previous->homopolymer()) {
string seq = referenceSequence.substr(prev_end_ref, indel.position - prev_end_ref);
string readseq = alternateSequence.substr(prev_end_read, indel.position - prev_end_ref);
VCFLEFTALIGN_DEBUG("seq: " << seq << endl << "readseq: " << readseq << endl);
if (previous->sequence.at(0) == seq.at(0)
&& FBhomopolymer(seq)
&& FBhomopolymer(readseq)) {
VCFLEFTALIGN_DEBUG("moving " << *previous << " right to "
<< (indel.insertion ? indel.position : indel.position - previous->length) << endl);
previous->position = indel.insertion ? indel.position : indel.position - previous->length;
}
}
else {
int pos = previous->position;
while (pos < (int) referenceSequence.length() &&
((previous->insertion && pos + previous->length <= indel.position)
||
(!previous->insertion && pos + previous->length < indel.position))
&& previous->sequence
== referenceSequence.substr(pos + previous->length, previous->length)) {
pos += previous->length;
}
if (pos < previous->position &&
((previous->insertion && pos + previous->length == indel.position)
||
(!previous->insertion && pos == indel.position - previous->length))
) {
VCFLEFTALIGN_DEBUG("right-merging tandem repeat: moving " << *previous << " right to " << pos << endl);
previous->position = pos;
}
}
}
previous = id;
}
}
// for each indel
// if ( we're matched up to the previous insertion (or deletion)
// and it's also an insertion or deletion )
// merge the indels
//
// and simultaneously reconstruct the cigar
Cigar newCigar;
if (!softBegin.empty()) {
newCigar.push_back(make_pair(softBegin.size(), "S"));
}
vector<VCFIndelAllele>::iterator id = indels.begin();
VCFIndelAllele last = *id++;
if (last.position > 0) {
newCigar.push_back(make_pair(last.position, "M"));
newCigar.push_back(make_pair(last.length, (last.insertion ? "I" : "D")));
} else {
newCigar.push_back(make_pair(last.length, (last.insertion ? "I" : "D")));
}
int lastend = last.insertion ? last.position : (last.position + last.length);
VCFLEFTALIGN_DEBUG(last << ",");
for (; id != indels.end(); ++id) {
VCFIndelAllele& indel = *id;
VCFLEFTALIGN_DEBUG(indel << ",");
if (indel.position < lastend) {
cerr << "impossibility?: indel realigned left of another indel" << endl
<< referenceSequence << endl << alternateSequence << endl;
exit(1);
} else if (indel.position == lastend && indel.insertion == last.insertion) {
pair<int, string>& op = newCigar.back();
op.first += indel.length;
} else if (indel.position >= lastend) { // also catches differential indels, but with the same position
newCigar.push_back(make_pair(indel.position - lastend, "M"));
newCigar.push_back(make_pair(indel.length, (indel.insertion ? "I" : "D")));
}
last = *id;
lastend = last.insertion ? last.position : (last.position + last.length);
}
if (lastend < alignedLength) {
newCigar.push_back(make_pair(alignedLength - lastend, "M"));
}
if (!softEnd.empty()) {
newCigar.push_back(make_pair(softEnd.size(), "S"));
}
VCFLEFTALIGN_DEBUG(endl);
cigar = newCigar;
for (vector<pair<int, string> >::const_iterator c = cigar.begin();
c != cigar.end(); ++c) {
unsigned int l = c->first;
char t = c->second.at(0);
cigar_after << l << t;
}
//cerr << cigar_before.str() << " changes to " << cigar_after.str() << endl;
VCFLEFTALIGN_DEBUG(cigar_after.str() << endl);
// check if we're realigned
if (cigar_after.str() == cigar_before.str()) {
return false;
} else {
return true;
}
}
TEST(CigarStringTest, ToStdString_Empty)
{
const string empty;
Cigar cigar;
EXPECT_EQ(empty, cigar.ToStdString());
}
TEST(CigarStringTest, FromStdString_Empty)
{
const string emptyCigar = "";
Cigar cigar = Cigar::FromStdString(emptyCigar);
EXPECT_TRUE(cigar.empty());
}
void BAMUtils::padded_alignment() {
Cigar cig = bam_record.get_cigar();
Sequence tdna = bam_record.get_seq();
int sdna_pos = 0;
int tdna_pos = 0;
pad_source.reserve(t_dna.length());
pad_target.reserve(t_dna.length());
pad_match.reserve(t_dna.length());
Sequence::iterator tdna_itr = tdna.get_iterator();
int tot = 0;
//find out if the first cigar op could be soft clipped or not
is_three_prime_soft_clipped = false;
for (Cigar::iterator i = cig.get_iterator(); i.good(); i.next()) {
//i.op(); i.len();
if (this->bam_record.mapped_reverse_strand()) {
if (tot > ( cig.get_length( ) - 3) ){
if (i.op() == 'S')
is_three_prime_soft_clipped = true;
else
is_three_prime_soft_clipped = false;
}
} else {
if (tot < 2) {
if (i.op() == 'S')
is_three_prime_soft_clipped = true;
else
is_three_prime_soft_clipped = false;
}
}
if (i.op() == 'I' ) {
pad_source.append(i.len(), '-');
int count = 0;
tdna_itr.set_position(tdna_pos);
while (tdna_itr.good()) {
if (count >= i.len()) {
break;
} else {
pad_target += tdna_itr.get();
tdna_itr.next();
tdna_pos++;
count++;
}
}
pad_match.append(i.len(), '+');
}
else if(i.op() == 'D' || i.op() == 'N') {
pad_source.append( t_dna.substr(sdna_pos, i.len()));
sdna_pos += i.len();
pad_target.append(i.len(), '-');
pad_match.append(i.len(), '-');
}
else if(i.op() == 'P') {
pad_source.append(i.len(), '*');
pad_target.append(i.len(), '*');
pad_match.append(i.len(), ' ');
} else if (i.op() == 'S') {
if (!truncate_soft_clipped) {
pad_source.append(i.len(), '-');
pad_match.append(i.len(), '+');
pad_target.append(i.len(), '+');
}
int count = 0;
while (tdna_itr.good()) {
if (count >= i.len()) {
break;
}
tdna_pos++;
tdna_itr.next();
count++;
}
}
else if (i.op() == 'H') {
//nothing for clipped bases
}else {
std::string ps, pt, pm;
ps.reserve(i.len());
pm.reserve(i.len());
ps = t_dna.substr(sdna_pos,i.len()); //tdna is really qdna
tdna_itr.set_position(tdna_pos);
int count = 0;
while (tdna_itr.good()) {
if (count < i.len()) {
pt += tdna_itr.get();
} else {
break;
}
tdna_itr.next();
count++;
}
for (unsigned int z = 0; z < ps.length(); z++) {
if (ps[z] == pt[z]) {
pad_match += '|';
} else if (ps[z] != 'A' || ps[z] != 'C' || ps[z] != 'G' || ps[z] != 'T') {
if (iupac_flag) {
std::vector<char> nukes(IUPAC::get_base(ps[z]));
bool replaced = false;
unsigned int nuke_ptr = 0;
for (unsigned int n = 0; n < nukes.size(); n++) {
if (nukes[n] == pt[z]) {
pad_match += '|';
replaced = true;
nuke_ptr = n;
break;
}
//nuke_ptr++;
}
if (!replaced) {
pad_match += ' ';
}
else if (!keep_iupac) {
//std::cerr << "nukes["<<nuke_ptr<<"]: " << nukes[nuke_ptr] << " nukes.size() " << nukes.size() << std::endl;
ps[z] = nukes[nuke_ptr];
}//keep_iupac
}//iupac_flag
else {
pad_match += ' ';
}
}//end else if checking ps[z] agianst nukes
else {
pad_match += ' ';
}
}//end for loop
pad_source += ps;
pad_target += pt;
sdna_pos += i.len();
tdna_pos += i.len();
}
tot++;
}
/*
std::cerr << "pad_source: " << pad_source << std::endl;
std::cerr << "pad_target: " << pad_target << std::endl;
std::cerr << "pad_match : " << pad_match << std::endl;
*/
}
void BAMUtils::dna() {
MD md = bam_record.get_md();
Cigar cig = bam_record.get_cigar();
Sequence qseq = bam_record.get_seq();
int position = 0;
std::string seq;
Sequence::iterator qseq_itr = qseq.get_iterator();
for (Cigar::iterator i = cig.get_iterator(); i.good(); i.next()) {
if (i.op() == 'M') {
int count = 0;
while (qseq_itr.good()) {
if (count >= i.len()) {
break;
} else {
seq += qseq_itr.get();
qseq_itr.next();
count++;
}
}
} else if ((i.op() == 'I') || (i.op() == 'S')) {
int count = 0;
while (qseq_itr.good()) {
if (count >= i.len()) {
break;
}
qseq_itr.next();
count++;
}
//bool is_error = false;
if (i.op() == 'S') {
soft_clipped_bases += i.len();
//is_error = true;
}
}
position++;
}
t_dna.reserve(seq.length());
int start = 0;
MD::iterator md_itr = md.get_iterator();
std::string num;
coord_t md_len = 0;
char cur;
while (md_itr.good()) {
cur = md_itr.get();
if (std::isdigit(cur)) {
num+=cur;
//md_itr.next();
}
else {
if (num.length() > 0) {
md_len = convert(num);
num.clear();
t_dna += seq.substr(start, md_len);
start += md_len;
}
}
if (cur == '^') {
//get nuc
md_itr.next();
char nuc = md_itr.get();
while (std::isalpha(nuc)) {
t_dna += nuc;
md_itr.next();
nuc = md_itr.get();
}
num += nuc; //it's a number now will
//lose this value if i don't do it here
//cur = nuc;
} else if (std::isalpha(cur)) {
t_dna += cur;
start++;
}
md_itr.next();
}
//clean up residual num if there is any
if (num.length() > 0) {
md_len = convert(num);
num.clear();
t_dna += seq.substr(start, md_len);
start += md_len;
}
}
int Stats::execute(int argc, char **argv)
{
// Extract command line arguments.
String inFile = "";
String indexFile = "";
bool basic = false;
bool noeof = false;
bool params = false;
bool qual = false;
bool phred = false;
int maxNumReads = -1;
bool unmapped = false;
String pBaseQC = "";
String cBaseQC = "";
String regionList = "";
int excludeFlags = 0;
int requiredFlags = 0;
bool withinRegion = false;
int minMapQual = 0;
String dbsnp = "";
PosList *dbsnpListPtr = NULL;
bool baseSum = false;
int bufferSize = PileupHelper::DEFAULT_WINDOW_SIZE;
ParameterList inputParameters;
BEGIN_LONG_PARAMETERS(longParameterList)
LONG_PARAMETER_GROUP("Required Parameters")
LONG_STRINGPARAMETER("in", &inFile)
LONG_PARAMETER_GROUP("Types of Statistics")
LONG_PARAMETER("basic", &basic)
LONG_PARAMETER("qual", &qual)
LONG_PARAMETER("phred", &phred)
LONG_STRINGPARAMETER("pBaseQC", &pBaseQC)
LONG_STRINGPARAMETER("cBaseQC", &cBaseQC)
LONG_PARAMETER_GROUP("Optional Parameters")
LONG_INTPARAMETER("maxNumReads", &maxNumReads)
LONG_PARAMETER("unmapped", &unmapped)
LONG_STRINGPARAMETER("bamIndex", &indexFile)
LONG_STRINGPARAMETER("regionList", ®ionList)
LONG_INTPARAMETER("excludeFlags", &excludeFlags)
LONG_INTPARAMETER("requiredFlags", &requiredFlags)
LONG_PARAMETER("noeof", &noeof)
LONG_PARAMETER("params", ¶ms)
LONG_PARAMETER_GROUP("Optional phred/qual Only Parameters")
LONG_PARAMETER("withinRegion", &withinRegion)
LONG_PARAMETER_GROUP("Optional BaseQC Only Parameters")
LONG_PARAMETER("baseSum", &baseSum)
LONG_INTPARAMETER("bufferSize", &bufferSize)
LONG_INTPARAMETER("minMapQual", &minMapQual)
LONG_STRINGPARAMETER("dbsnp", &dbsnp)
END_LONG_PARAMETERS();
inputParameters.Add(new LongParameters ("Input Parameters",
longParameterList));
inputParameters.Read(argc-1, &(argv[1]));
// If no eof block is required for a bgzf file, set the bgzf file type to
// not look for it.
if(noeof)
{
// Set that the eof block is not required.
BgzfFileType::setRequireEofBlock(false);
}
// Check to see if the in file was specified, if not, report an error.
if(inFile == "")
{
usage();
inputParameters.Status();
// In file was not specified but it is mandatory.
std::cerr << "--in is a mandatory argument for stats, "
<< "but was not specified" << std::endl;
return(-1);
}
// Use the index file if unmapped or regionList is not empty.
bool useIndex = (unmapped|| (!regionList.IsEmpty()));
// IndexFile is required, so check to see if it has been set.
if(useIndex && (indexFile == ""))
{
// In file was not specified, so set it to the in file
// + ".bai"
indexFile = inFile + ".bai";
}
////////////////////////////////////////
// Setup in case pileup is used.
Pileup<PileupElementBaseQCStats> pileup(bufferSize);
// Initialize start/end positions.
myStartPos = 0;
myEndPos = -1;
// Open the output qc file if applicable.
IFILE baseQCPtr = NULL;
if(!pBaseQC.IsEmpty() && !cBaseQC.IsEmpty())
{
usage();
inputParameters.Status();
// Cannot specify both types of baseQC.
std::cerr << "Cannot specify both --pBaseQC & --cBaseQC." << std::endl;
return(-1);
}
else if(!pBaseQC.IsEmpty())
{
baseQCPtr = ifopen(pBaseQC, "w");
PileupElementBaseQCStats::setPercentStats(true);
}
else if(!cBaseQC.IsEmpty())
{
baseQCPtr = ifopen(cBaseQC, "w");
PileupElementBaseQCStats::setPercentStats(false);
}
if(baseQCPtr != NULL)
{
PileupElementBaseQCStats::setOutputFile(baseQCPtr);
PileupElementBaseQCStats::printHeader();
}
if((baseQCPtr != NULL) || baseSum)
{
PileupElementBaseQCStats::setMapQualFilter(minMapQual);
PileupElementBaseQCStats::setBaseSum(baseSum);
}
if(params)
{
inputParameters.Status();
}
// Open the file for reading.
SamFile samIn;
if(!samIn.OpenForRead(inFile))
{
fprintf(stderr, "%s\n", samIn.GetStatusMessage());
return(samIn.GetStatus());
}
samIn.SetReadFlags(requiredFlags, excludeFlags);
// Set whether or not basic statistics should be generated.
samIn.GenerateStatistics(basic);
// Read the sam header.
SamFileHeader samHeader;
if(!samIn.ReadHeader(samHeader))
{
fprintf(stderr, "%s\n", samIn.GetStatusMessage());
return(samIn.GetStatus());
}
// Open the bam index file for reading if we are
// doing unmapped reads (also set the read section).
if(useIndex)
{
samIn.ReadBamIndex(indexFile);
if(unmapped)
{
samIn.SetReadSection(-1);
}
if(!regionList.IsEmpty())
{
myRegionList = ifopen(regionList, "r");
}
}
//////////////////////////
// Read dbsnp if specified and doing baseQC
if(((baseQCPtr != NULL) || baseSum) && (!dbsnp.IsEmpty()))
{
// Read the dbsnp file.
IFILE fdbSnp;
fdbSnp = ifopen(dbsnp,"r");
// Determine how many entries.
const SamReferenceInfo& refInfo = samHeader.getReferenceInfo();
int maxRefLen = 0;
for(int i = 0; i < refInfo.getNumEntries(); i++)
{
int refLen = refInfo.getReferenceLength(i);
if(refLen >= maxRefLen)
{
maxRefLen = refLen + 1;
}
}
dbsnpListPtr = new PosList(refInfo.getNumEntries(),maxRefLen);
if(fdbSnp==NULL)
{
std::cerr << "Open dbSNP file " << dbsnp.c_str() << " failed!\n";
}
else if(dbsnpListPtr == NULL)
{
std::cerr << "Failed to init the memory allocation for the dbsnpList.\n";
}
else
{
// Read the dbsnp file.
StringArray tokens;
String buffer;
int position = 0;
int refID = 0;
// Loop til the end of the file.
while (!ifeof(fdbSnp))
{
// Read the next line.
buffer.ReadLine(fdbSnp);
// If it does not have at least 2 columns,
// continue to the next line.
if (buffer.IsEmpty() || buffer[0] == '#') continue;
tokens.AddTokens(buffer);
if(tokens.Length() < 2) continue;
if(!tokens[1].AsInteger(position))
{
std::cerr << "Improperly formatted region line, start position "
<< "(2nd column) is not an integer: "
<< tokens[1]
<< "; Skipping to the next line.\n";
continue;
}
// Look up the reference name.
refID = samHeader.getReferenceID(tokens[0]);
if(refID != SamReferenceInfo::NO_REF_ID)
{
// Reference id was found, so add it to the dbsnp
dbsnpListPtr->addPosition(refID, position);
}
tokens.Clear();
buffer.Clear();
}
}
ifclose(fdbSnp);
}
// Read the sam records.
SamRecord samRecord;
int numReads = 0;
//////////////////////
// Setup in case doing a quality count.
// Quality histogram.
const int MAX_QUAL = 126;
const int START_QUAL = 33;
uint64_t qualCount[MAX_QUAL+1];
for(int i = 0; i <= MAX_QUAL; i++)
{
qualCount[i] = 0;
}
const int START_PHRED = 0;
const int PHRED_DIFF = START_QUAL - START_PHRED;
const int MAX_PHRED = MAX_QUAL - PHRED_DIFF;
uint64_t phredCount[MAX_PHRED+1];
for(int i = 0; i <= MAX_PHRED; i++)
{
phredCount[i] = 0;
}
int refPos = 0;
Cigar* cigarPtr = NULL;
char cigarChar = '?';
// Exclude clips from the qual/phred counts if unmapped reads are excluded.
bool qualExcludeClips = excludeFlags & SamFlag::UNMAPPED;
//////////////////////////////////
// When not reading by sections, getNextSection returns true
// the first time, then false the next time.
while(getNextSection(samIn))
{
// Keep reading records from the file until SamFile::ReadRecord
// indicates to stop (returns false).
while(((maxNumReads < 0) || (numReads < maxNumReads)) && samIn.ReadRecord(samHeader, samRecord))
{
// Another record was read, so increment the number of reads.
++numReads;
// See if the quality histogram should be genereated.
if(qual || phred)
{
// Get the quality.
const char* qual = samRecord.getQuality();
// Check for no quality ('*').
if((qual[0] == '*') && (qual[1] == 0))
{
// This record does not have a quality string, so no
// quality processing is necessary.
}
else
{
int index = 0;
cigarPtr = samRecord.getCigarInfo();
cigarChar = '?';
refPos = samRecord.get0BasedPosition();
if(!qualExcludeClips && (cigarPtr != NULL))
{
// Offset the reference position by any soft clips
// by subtracting the queryIndex of this start position.
// refPos is now the start position of the clips.
refPos -= cigarPtr->getQueryIndex(0);
}
while(qual[index] != 0)
{
// Skip this quality if it is clipped and we are skipping clips.
if(cigarPtr != NULL)
{
cigarChar = cigarPtr->getCigarCharOpFromQueryIndex(index);
}
if(qualExcludeClips && Cigar::isClip(cigarChar))
{
// Skip a clipped quality.
++index;
// Increment the position.
continue;
}
if(withinRegion && (myEndPos != -1) && (refPos >= myEndPos))
{
// We have hit the end of the region, stop processing this
// quality string.
break;
}
if(withinRegion && (refPos < myStartPos))
{
// This position is not in the target.
++index;
// Update the position if this is found in the reference or a clip.
if(Cigar::foundInReference(cigarChar) || Cigar::isClip(cigarChar))
{
++refPos;
}
continue;
}
// Check for valid quality.
if((qual[index] < START_QUAL) || (qual[index] > MAX_QUAL))
{
if(qual)
{
std::cerr << "Invalid Quality found: " << qual[index]
<< ". Must be between "
<< START_QUAL << " and " << MAX_QUAL << ".\n";
}
if(phred)
{
std::cerr << "Invalid Phred Quality found: " << qual[index] - PHRED_DIFF
<< ". Must be between "
<< START_QUAL << " and " << MAX_QUAL << ".\n";
}
// Skip an invalid quality.
++index;
// Update the position if this is found in the reference or a clip.
if(Cigar::foundInReference(cigarChar) || Cigar::isClip(cigarChar))
{
++refPos;
}
continue;
}
// Increment the count for this quality.
++(qualCount[(int)(qual[index])]);
++(phredCount[(int)(qual[index]) - PHRED_DIFF]);
// Update the position if this is found in the reference or a clip.
if(Cigar::foundInReference(cigarChar) || Cigar::isClip(cigarChar))
{
++refPos;
}
++index;
}
}
}
// Check the next thing to do for the read.
if((baseQCPtr != NULL) || baseSum)
{
// Pileup the bases for this read.
pileup.processAlignmentRegion(samRecord, myStartPos, myEndPos, dbsnpListPtr);
}
}
// Done with a section, move on to the next one.
// New section, so flush the pileup.
pileup.flushPileup();
}
// Flush the rest of the pileup.
if((baseQCPtr != NULL) || baseSum)
{
// Pileup the bases.
pileup.processAlignmentRegion(samRecord, myStartPos, myEndPos, dbsnpListPtr);
PileupElementBaseQCStats::printSummary();
ifclose(baseQCPtr);
}
std::cerr << "Number of records read = " <<
samIn.GetCurrentRecordCount() << std::endl;
if(basic)
{
std::cerr << std::endl;
samIn.PrintStatistics();
}
// Print the quality stats.
if(qual)
{
std::cerr << std::endl;
std::cerr << "Quality\tCount\n";
for(int i = START_QUAL; i <= MAX_QUAL; i++)
{
std::cerr << i << "\t" << qualCount[i] << std::endl;
}
}
// Print the phred quality stats.
if(phred)
{
std::cerr << std::endl;
std::cerr << "Phred\tCount\n";
for(int i = START_PHRED; i <= MAX_PHRED; i++)
{
std::cerr << i << "\t" << phredCount[i] << std::endl;
}
}
SamStatus::Status status = samIn.GetStatus();
if(status == SamStatus::NO_MORE_RECS)
{
// A status of NO_MORE_RECS means that all reads were successful.
status = SamStatus::SUCCESS;
}
return(status);
}
// Soft clip the cigar from the front and/or the back, writing the value
// into the new cigar.
SamFilter::FilterStatus SamFilter::softClip(Cigar& oldCigar,
int32_t numFrontClips,
int32_t numBackClips,
int32_t& startPos,
CigarRoller& updatedCigar)
{
int32_t readLength = oldCigar.getExpectedQueryBaseCount();
int32_t endClipPos = readLength - numBackClips;
FilterStatus status = NONE;
if((numFrontClips != 0) || (numBackClips != 0))
{
// Clipping from front and/or from the back.
// Check to see if the entire read was clipped.
int32_t totalClips = numFrontClips + numBackClips;
if(totalClips >= readLength)
{
/////////////////////////////
// The entire read is clipped, so rather than clipping it,
// filter it out.
return(FILTERED);
}
// Part of the read was clipped.
status = CLIPPED;
// Loop through, creating an updated cigar.
int origCigarOpIndex = 0;
// Track how many read positions are covered up to this
// point by the cigar to determine up to up to what
// point in the cigar is affected by this clipping.
int32_t numPositions = 0;
// Track if any non-clips are in the new cigar.
bool onlyClips = true;
const Cigar::CigarOperator* op = NULL;
//////////////////
// Clip from front
while((origCigarOpIndex < oldCigar.size()) &&
(numPositions < numFrontClips))
{
op = &(oldCigar.getOperator(origCigarOpIndex));
switch(op->operation)
{
case Cigar::hardClip:
// Keep this operation as the new clips do not
// affect other clips.
updatedCigar += *op;
break;
case Cigar::del:
case Cigar::skip:
// Skip and delete are going to be dropped, and
// are not in the read, so the read index doesn't
// need to be updated
break;
case Cigar::insert:
case Cigar::match:
case Cigar::mismatch:
case Cigar::softClip:
// Update the read index as these types
// are found in the read.
numPositions += op->count;
break;
case Cigar::none:
default:
// Nothing to do for none.
break;
};
++origCigarOpIndex;
}
// If bases were clipped from the front, add the clip and
// any partial cigar operation as necessary.
if(numFrontClips != 0)
{
// Add the softclip to the front of the read.
updatedCigar.Add(Cigar::softClip, numFrontClips);
// Add the rest of the last Cigar operation if
// it is not entirely clipped.
int32_t newCount = numPositions - numFrontClips;
if(newCount > 0)
{
// Before adding it, check to see if the same
// operation is clipped from the end.
// numPositions greater than the endClipPos
// means that it is equal or past that position,
// so shorten the number of positions.
if(numPositions > endClipPos)
{
newCount -= (numPositions - endClipPos);
}
if(newCount > 0)
{
updatedCigar.Add(op->operation, newCount);
if(!Cigar::isClip(op->operation))
{
onlyClips = false;
}
}
}
}
// Add operations until the point of the end clip is reached.
// For example...
// 2M1D3M = MMDMMM readLength = 5
// readIndex: 01 234
// at cigarOpIndex 0 (2M), numPositions = 2.
// at cigarOpIndex 1 (1D), numPositions = 2.
// at cigarOpIndex 2 (3M), numPositions = 5.
// if endClipPos = 2, we still want to consume the 1D, so
// need to keep looping until numPositions > endClipPos
while((origCigarOpIndex < oldCigar.size()) &&
(numPositions <= endClipPos))
{
op = &(oldCigar.getOperator(origCigarOpIndex));
// Update the numPositions count if the operations indicates
// bases within the read.
if(!Cigar::foundInQuery(op->operation))
{
// This operation is not in the query read sequence,
// so it is not yet to the endClipPos, just add the
// operation do not increment the number of positions.
updatedCigar += *op;
if(!Cigar::isClip(op->operation))
{
onlyClips = false;
}
}
else
{
// This operation appears in the query sequence, so
// check to see if the clip occurs in this operation.
// endClipPos is 0 based & numPositions is a count.
// If endClipPos is 4, then it is the 5th position.
// If 4 positions are covered so far (numPositions = 4),
// then we are right at endCLipPos: 4-4 = 0, none of
// this operation should be kept.
// If only 3 positions were covered, then we are at offset
// 3, so offset 3 should be added: 4-3 = 1.
uint32_t numPosTilClip = endClipPos - numPositions;
if(numPosTilClip < op->count)
{
// this operation is partially clipped, write the part
// that was not clipped if it is not all clipped.
if(numPosTilClip != 0)
{
updatedCigar.Add(op->operation,
numPosTilClip);
if(!Cigar::isClip(op->operation))
{
onlyClips = false;
}
}
}
else
{
// This operation is not clipped, so add it
updatedCigar += *op;
if(!Cigar::isClip(op->operation))
{
onlyClips = false;
}
}
// This operation occurs in the query sequence, so
// increment the number of positions covered.
numPositions += op->count;
}
// Move to the next cigar position.
++origCigarOpIndex;
}
//////////////////
// Add the softclip to the back.
if(numBackClips != 0)
{
// Add the softclip to the end
updatedCigar.Add(Cigar::softClip, numBackClips);
}
//////////////////
// Add any hardclips remaining in the original cigar to the back.
while(origCigarOpIndex < oldCigar.size())
{
op = &(oldCigar.getOperator(origCigarOpIndex));
if(op->operation == Cigar::hardClip)
{
// Keep this operation as the new clips do not
// affect other clips.
updatedCigar += *op;
}
++origCigarOpIndex;
}
// Check to see if the new cigar is only clips.
if(onlyClips)
{
// Only clips in the new cigar, so mark the read as filtered
// instead of updating the cigar.
/////////////////////////////
// The entire read was clipped.
status = FILTERED;
}
else
{
// Part of the read was clipped.
// Update the starting position if a clip was added to
// the front.
if(numFrontClips > 0)
{
// Convert from query index to reference position (from the
// old cigar)
// Get the position for the last front clipped position by
// getting the position associated with the clipped base on
// the reference. Then add one to get to the first
// non-clipped position.
int32_t lastFrontClipPos = numFrontClips - 1;
int32_t newStartPos = oldCigar.getRefPosition(lastFrontClipPos,
startPos);
if(newStartPos != Cigar::INDEX_NA)
{
// Add one to get first non-clipped position.
startPos = newStartPos + 1;
}
}
}
}
return(status);
}
// Soft Clip from the beginning of the read to the specified reference position.
int32_t CigarHelper::softClipBeginByRefPos(SamRecord& record,
int32_t refPosition0Based,
CigarRoller& newCigar,
int32_t &new0BasedPosition)
{
newCigar.clear();
Cigar* cigar = record.getCigarInfo();
if(cigar == NULL)
{
// Failed to get the cigar.
ErrorHandler::handleError("Soft clipping, but failed to read the cigar");
return(NO_CLIP);
}
// No cigar or position in the record, so return no clip.
if((cigar->size() == 0) || (record.get0BasedPosition() == -1))
{
return(NO_CLIP);
}
// Check to see if the reference position occurs before the record starts,
// if it does, do no clipping.
if(refPosition0Based < record.get0BasedPosition())
{
// Not within this read, so nothing to clip.
newCigar.Set(record.getCigar());
return(NO_CLIP);
}
// The position falls after the read starts, so loop through until the
// position or the end of the read is found.
int32_t readClipPosition = 0;
bool clipWritten = false;
new0BasedPosition = record.get0BasedPosition();
for(int i = 0; i < cigar->size(); i++)
{
const Cigar::CigarOperator* op = &(cigar->getOperator(i));
if(clipWritten)
{
// Clip point has been found, so just add everything.
newCigar += *op;
// Go to the next operation.
continue;
}
// The clip point has not yet been found, so check to see if we found
// it now.
// Not a clip, check to see if the operation is found in the
// reference.
if(Cigar::foundInReference(*op))
{
// match, mismatch, deletion, skip
// increment the current reference position to just past this
// operation.
new0BasedPosition += op->count;
// Check to see if this is also in the query, because otherwise
// the operation is still being consumed.
if(Cigar::foundInQuery(*op))
{
// Also in the query, determine if the entire thing should
// be clipped or just part of it.
uint32_t numKeep = 0;
// Check to see if we have hit our clip position.
if(refPosition0Based < new0BasedPosition)
{
// The specified clip position is in this cigar operation.
numKeep = new0BasedPosition - refPosition0Based - 1;
if(numKeep > op->count)
{
// Keep the entire read. This happens because
// we keep reading until the first match/mismatch
// after the clip.
numKeep = op->count;
}
}
// Add the part of this operation that is being clipped
// to the clip count.
readClipPosition += (op->count - numKeep);
// Only write the clip if we found a match/mismatch
// to write. Otherwise we will keep accumulating clips
// for the case of insertions.
if(numKeep > 0)
{
new0BasedPosition -= numKeep;
newCigar.Add(Cigar::softClip, readClipPosition);
// Add the clipped part of this cigar to the clip
// position.
newCigar.Add(op->operation, numKeep);
// Found a match after the clip point, so stop
// consuming cigar operations.
clipWritten = true;
continue;
}
}
}
else
{
// Only add hard clips. The softclips will be added in
// when the total number is found.
if(op->operation == Cigar::hardClip)
{
// Check if this is the first operation, if so, just write it.
if(i == 0)
{
newCigar += *op;
}
// Check if it is the last operation (otherwise skip it).
else if(i == (cigar->size() - 1))
{
// Check whether or not the clip was ever written, and if
// not, write it.
if(clipWritten == false)
{
newCigar.Add(Cigar::softClip, readClipPosition);
// Since no match/mismatch was ever found, set
// the new ref position to the original one.
new0BasedPosition = record.get0BasedPosition();
clipWritten = true;
}
// Add the hard clip.
newCigar += *op;
}
}
// Not yet to the clip position, so do not add this operation.
if(Cigar::foundInQuery(*op))
{
// Found in the query, so update the read clip position.
readClipPosition += op->count;
}
}
} // End loop through cigar.
// Check whether or not the clip was ever written, and if
// not, write it.
if(clipWritten == false)
{
newCigar.Add(Cigar::softClip, readClipPosition);
// Since no match/mismatch was ever found, set
// the new ref position to the original one.
new0BasedPosition = record.get0BasedPosition();
}
// Subtract 1 since readClipPosition atually contains the first 0based
// position that is not clipped.
return(readClipPosition - 1);
}
// Soft Clip from the end of the read at the specified reference position.
int32_t CigarHelper::softClipEndByRefPos(SamRecord& record,
int32_t refPosition0Based,
CigarRoller& newCigar)
{
newCigar.clear();
Cigar* cigar = record.getCigarInfo();
if(cigar == NULL)
{
// Failed to get the cigar.
ErrorHandler::handleError("Soft clipping, but failed to read the cigar");
return(NO_CLIP);
}
// No cigar or position in the record, so return no clip.
if((cigar->size() == 0) || (record.get0BasedPosition() == -1))
{
return(NO_CLIP);
}
// Check to see if the reference position occurs after the record ends,
// if so, do no clipping.
if(refPosition0Based > record.get0BasedAlignmentEnd())
{
// Not within this read, so nothing to clip.
newCigar.Set(record.getCigar());
return(NO_CLIP);
}
// The position falls before the read ends, so loop through until the
// position is found.
int32_t currentRefPosition = record.get0BasedPosition();
int32_t readClipPosition = 0;
for(int i = 0; i < cigar->size(); i++)
{
const Cigar::CigarOperator* op = &(cigar->getOperator(i));
// If the operation is found in the reference, increase the
// reference position.
if(Cigar::foundInReference(*op))
{
// match, mismatch, deletion, skip
// increment the current reference position to just past
// this operation.
currentRefPosition += op->count;
}
// Check to see if we have hit our clip position.
if(refPosition0Based < currentRefPosition)
{
// If this read is also in the query (match/mismatch),
// write the partial op to the new cigar.
int32_t numKeep = 0;
if(Cigar::foundInQuery(*op))
{
numKeep = op->count - (currentRefPosition - refPosition0Based);
if(numKeep > 0)
{
newCigar.Add(op->operation, numKeep);
readClipPosition += numKeep;
}
}
else if(Cigar::isClip(*op))
{
// This is a hard clip, so write it.
newCigar.Add(op->operation, op->count);
}
else
{
// Not found in the query (skip/deletion),
// so don't write any of the operation.
}
// Found the clip point, so break.
break;
}
else if(refPosition0Based == currentRefPosition)
{
newCigar += *op;
if(Cigar::foundInQuery(*op))
{
readClipPosition += op->count;
}
}
else
{
// Not yet to the clip position, so add this operation/size to
// the new cigar.
newCigar += *op;
if(Cigar::foundInQuery(*op))
{
// Found in the query, so update the read clip position.
readClipPosition += op->count;
}
}
} // End loop through cigar.
// Before adding the softclip, read from the end of the cigar checking to
// see if the operations are in the query, removing operations that are
// not (pad/delete/skip) until a hardclip or an operation in the query is
// found. We do not want a pad/delete/skip right before a softclip.
for(int j = newCigar.size() - 1; j >= 0; j--)
{
const Cigar::CigarOperator* op = &(newCigar.getOperator(j));
if(!Cigar::foundInQuery(*op) && !Cigar::isClip(*op))
{
// pad/delete/skip
newCigar.Remove(j);
}
else if(Cigar::foundInQuery(*op) & Cigar::isClip(*op))
{
// Soft clip, so increment the clip position for the return value.
// Remove the softclip since the readClipPosition is used to
// calculate teh size of the soft clip added.
readClipPosition -= op->count;
newCigar.Remove(j);
}
else
{
// Found a cigar operation that should not be deleted, so stop deleting.
break;
}
}
// Determine the number of soft clips.
int32_t numSoftClips = record.getReadLength() - readClipPosition;
// NOTE that if the previous operation is a softclip, the CigarRoller logic
// will merge this with that one.
newCigar.Add(Cigar::softClip, numSoftClips);
// Check if an ending hard clip needs to be added.
if(cigar->size() != 0)
{
const Cigar::CigarOperator* lastOp =
&(cigar->getOperator(cigar->size() - 1));
if(lastOp->operation == Cigar::hardClip)
{
newCigar += *lastOp;
}
}
return(readClipPosition);
}
std::string Cigar::toString(const Cigar &cigarBuffer, unsigned offset, unsigned length)
{
ISAAC_ASSERT_MSG(cigarBuffer.size() >= offset + length, "Requested end is outside of cigarBuffer");
return toString(cigarBuffer.begin() + offset, cigarBuffer.begin() + offset + length);
}
// Add an entry to this pileup element.
void PileupElementBaseQual::addEntry(SamRecord& record)
{
// Call the base class:
PileupElement::addEntry(record);
if(myRefAllele.empty())
{
genomeIndex_t markerIndex = (*myRefSeq).getGenomePosition(getChromosome(), static_cast<uint32_t>(getRefPosition()+1));
myRefAllele = (*myRefSeq)[markerIndex];
}
// Increment the index
++myIndex;
// if the index has gone beyond the allocated space, double the size.
if(myIndex >= myAllocatedSize)
{
char* tempBuffer = (char*)realloc(myBases, myAllocatedSize * 2);
if(tempBuffer == NULL)
{
std::cerr << "Memory Allocation Failure\n";
// TODO
return;
}
myBases = tempBuffer;
int8_t* tempInt8Buffer = (int8_t*)realloc(myMapQualities, myAllocatedSize * 2 * sizeof(int8_t));
if(tempInt8Buffer == NULL)
{
std::cerr << "Memory Allocation Failure\n";
// TODO
return;
}
myMapQualities = tempInt8Buffer;
tempInt8Buffer = (int8_t*)realloc(myQualities, myAllocatedSize * 2 * sizeof(int8_t));
if(tempInt8Buffer == NULL)
{
std::cerr << "Memory Allocation Failure\n";
// TODO
return;
}
myQualities = tempInt8Buffer;
tempBuffer = (char*)realloc(myStrands, myAllocatedSize * 2);
if(tempBuffer == NULL)
{
std::cerr << "Memory Allocation Failure\n";
// TODO
return;
}
myStrands = tempBuffer;
tempInt8Buffer = (int8_t*)realloc(myCycles, myAllocatedSize * 2 * sizeof(int8_t));
if(tempInt8Buffer == NULL)
{
std::cerr << "Memory Allocation Failure\n";
// TODO
return;
}
myCycles = tempInt8Buffer;
int16_t* tempInt16Buffer = (int16_t*)realloc(myGLScores, myAllocatedSize * 2 * sizeof(int16_t));
if(tempInt8Buffer == NULL)
{
std::cerr << "Memory Allocation Failure\n";
// TODO
return;
}
myGLScores = tempInt16Buffer;
myAllocatedSize = myAllocatedSize * 2;
}
Cigar* cigar = record.getCigarInfo();
if(cigar == NULL)
{
throw std::runtime_error("Failed to retrieve cigar info from the record.");
}
int32_t readIndex =
cigar->getQueryIndex(getRefPosition(), record.get0BasedPosition());
// If the readPosition is N/A, this is a deletion.
if(readIndex != CigarRoller::INDEX_NA)
{
char base = record.getSequence(readIndex);
int8_t mapQual = record.getMapQuality();
//-33 to obtain the PHRED base quality
char qual = record.getQuality(readIndex) - 33;
if(qual == UNSET_QUAL)
{
qual = ' ';
}
char strand = (record.getFlag() & 0x0010) ? 'R' : 'F';
int cycle = strand == 'F' ? readIndex + 1 : record.getReadLength() - readIndex;
myBases[myIndex] = base;
myMapQualities[myIndex] = mapQual;
myQualities[myIndex] = qual;
myStrands[myIndex] = strand;
myCycles[myIndex] = cycle;
}
else if(myAddDelAsBase)
{
int8_t mapQual = record.getMapQuality();
char strand = (record.getFlag() & 0x0010) ? 'R' : 'F';
myBases[myIndex] = '-';
myMapQualities[myIndex] = mapQual;
myQualities[myIndex] = -1;
myStrands[myIndex] = strand;
myCycles[myIndex] = -1;
}
else
{
// Do not add a deletion.
// Did not add any entries, so decrement the index counter since the
// index was not used.
--myIndex;
}
}
bool Recab::processReadBuildTable(SamRecord& samRecord)
{
static BaseData data;
static std::string chromosomeName;
static std::string readGroup;
static std::string aligTypes;
int seqLen = samRecord.getReadLength();
// Check if the parameters have been processed.
if(!myParamsSetup)
{
// This throws an exception if the reference cannot be setup.
processParams();
}
uint16_t flag = samRecord.getFlag();
if(!SamFlag::isMapped(flag))
{
// Unmapped, skip processing
++myUnMappedCount;
}
else
{
// This read is mapped.
++myMappedCount;
}
if(SamFlag::isSecondary(flag))
{
// Secondary read
++mySecondaryCount;
}
if(SamFlag::isDuplicate(flag))
{
++myDupCount;
}
if(SamFlag::isQCFailure(flag))
{
++myQCFailCount;
}
// Check if the flag contains an exclude.
if((flag & myIntBuildExcludeFlags) != 0)
{
// Do not use this read for building the recalibration table.
++myNumBuildSkipped;
return(false);
}
if(samRecord.getMapQuality() == 0)
{
// 0 mapping quality, so skip processing.
++myMapQual0Count;
++myNumBuildSkipped;
return(false);
}
if(samRecord.getMapQuality() == 255)
{
// 255 mapping quality, so skip processing.
++myMapQual255Count;
++myNumBuildSkipped;
return(false);
}
chromosomeName = samRecord.getReferenceName();
readGroup = samRecord.getString("RG").c_str();
// Look for the read group in the map.
// TODO - extra string constructor??
RgInsertReturn insertRet =
myRg2Id.insert(std::pair<std::string, uint16_t>(readGroup, 0));
if(insertRet.second == true)
{
// New element inserted.
insertRet.first->second = myId2Rg.size();
myId2Rg.push_back(readGroup);
}
data.rgid = insertRet.first->second;
//reverse
bool reverse;
if(SamFlag::isReverse(flag))
reverse = true;
else
reverse = false;
if(myReferenceGenome == NULL)
{
throw std::runtime_error("Failed to setup Reference File.\n");
}
genomeIndex_t mapPos =
myReferenceGenome->getGenomePosition(chromosomeName.c_str(),
samRecord.get1BasedPosition());
if(mapPos==INVALID_GENOME_INDEX)
{
Logger::gLogger->warning("INVALID_GENOME_INDEX (chrom:pos %s:%ld) and record skipped... Reference in BAM is different from the ref used here!", chromosomeName.c_str(), samRecord.get1BasedPosition());
++myNumBuildSkipped;
return false;
}
if(!myQField.IsEmpty())
{
// Check if there is an old quality.
const String* oldQPtr =
samRecord.getStringTag(myQField.c_str());
if((oldQPtr != NULL) && (oldQPtr->Length() == seqLen))
{
// There is an old quality, so use that.
myQualityStrings.oldq = oldQPtr->c_str();
}
else
{
// Tag was not found, so use the current quality.
++myNumQualTagErrors;
if(myNumQualTagErrors == 1)
{
Logger::gLogger->warning("Recab: %s tag was not found/invalid, so using the quality field in records without the tag", myQField.c_str());
}
myQualityStrings.oldq = samRecord.getQuality();
}
//printf("%s\n",samRecord.getQuality());
//printf("%s:%s\n",myQField.c_str(),temp.c_str());
}
else
{
myQualityStrings.oldq = samRecord.getQuality();
}
if(myQualityStrings.oldq.length() != (unsigned int)seqLen)
{
Logger::gLogger->warning("Quality is not the correct length, so skipping recalibration on that record.");
++myNumBuildSkipped;
return(false);
}
aligTypes = "";
Cigar* cigarPtr = samRecord.getCigarInfo();
if(cigarPtr == NULL)
{
Logger::gLogger->warning("Failed to get the cigar");
++myNumBuildSkipped;
return(false);
}
// This read will be used for building the recab table.
++myNumBuildReads;
////////////////
////// iterate sequence
////////////////
genomeIndex_t refPos = 0;
int32_t refOffset = 0;
int32_t prevRefOffset = Cigar::INDEX_NA;
int32_t seqPos = 0;
int seqIncr = 1;
if(reverse)
{
seqPos = seqLen - 1;
seqIncr = -1;
}
// read
if(!SamFlag::isPaired(flag) || SamFlag::isFirstFragment(flag))
// Mark as first if it is not paired or if it is the
// first in the pair.
data.read = 0;
else
data.read = 1;
// Set unsetbase for curBase.
// This will be used for the prebase of cycle 0.
data.curBase = 'K';
for (data.cycle = 0; data.cycle < seqLen; data.cycle++, seqPos += seqIncr)
{
// Store the previous current base in preBase.
data.preBase = data.curBase;
// Get the current base before checking if we are going to
// process this position so it will be set for the next position.
data.curBase = samRecord.getSequence(seqPos);
if(reverse)
{
// Complement the current base.
// The prebase is already complemented.
data.curBase =
BaseAsciiMap::base2complement[(unsigned int)(data.curBase)];
}
// Get the reference offset.
refOffset = cigarPtr->getRefOffset(seqPos);
if(refOffset == Cigar::INDEX_NA)
{
// Not a match/mismatch, so continue to the next one which will
// not have a previous match/mismatch.
// Set previous ref offset to a negative so
// the next one won't be kept.
prevRefOffset = -2;
continue;
}
// This one is a match.
refPos = mapPos + refOffset;
// Check to see if we should process this position.
// Do not process if it is cycle 0 and:
// 1) current base is in dbsnp
if(data.cycle == 0)
{
if(!(myDbsnpFile.IsEmpty()) && myDbSNP[refPos])
{
// Save the previous reference offset.
++myNumDBSnpSkips;
prevRefOffset = refOffset;
continue;
}
}
else
{
// Do not process if it is not cycle 0 and:
// 1) previous reference position not adjacent
// (not a match/mismatch)
// 2) previous base is in dbsnp
// 3) current base is in dbsnp
if((!myKeepPrevNonAdjacent && (refOffset != (prevRefOffset + seqIncr))) ||
(data.preBase == 'K'))
{
// Save the previous reference offset.
prevRefOffset = refOffset;
continue;
}
if(!(myDbsnpFile.IsEmpty()) &&
(myDbSNP[refPos] ||
(!myKeepPrevDbsnp && myDbSNP[refPos - seqIncr])))
{
++myNumDBSnpSkips;
// Save the previous reference offset.
prevRefOffset = refOffset;
continue;
}
}
// Save the previous reference offset.
prevRefOffset = refOffset;
// Set the reference & read bases in the Covariates
char refBase = (*myReferenceGenome)[refPos];
if(BaseUtilities::isAmbiguous(refBase))
{
// N reference, so skip it when building the table.
++myAmbiguous;
continue;
}
if(reverse)
{
refBase = BaseAsciiMap::base2complement[(unsigned int)(refBase)];
}
// Get quality char
data.qual =
BaseUtilities::getPhredBaseQuality(myQualityStrings.oldq[seqPos]);
// skip bases with quality below the minimum set.
if(data.qual < myMinBaseQual)
{
++mySubMinQual;
continue;
}
if(BaseUtilities::areEqual(refBase, data.curBase)
&& (BaseAsciiMap::base2int[(unsigned int)(data.curBase)] < 4))
myBMatchCount++;
else
myBMismatchCount++;
hasherrormodel.setCell(data, refBase);
myBasecounts++;
}
return true;
}
void realign_bam(Parameters& params) {
FastaReference reference;
reference.open(params.fasta_reference);
bool suppress_output = false;
int dag_window_size = params.dag_window_size;
// open BAM file
BamReader reader;
if (!reader.Open("stdin")) {
cerr << "could not open stdin for reading" << endl;
exit(1);
}
BamWriter writer;
if (!params.dry_run && !writer.Open("stdout", reader.GetHeaderText(), reader.GetReferenceData())) {
cerr << "could not open stdout for writing" << endl;
exit(1);
}
// store the names of all the reference sequences in the BAM file
map<int, string> referenceIDToName;
vector<RefData> referenceSequences = reader.GetReferenceData();
int i = 0;
for (RefVector::iterator r = referenceSequences.begin(); r != referenceSequences.end(); ++r) {
referenceIDToName[i] = r->RefName;
++i;
}
vcf::VariantCallFile vcffile;
if (!params.vcf_file.empty()) {
if (!vcffile.open(params.vcf_file)) {
cerr << "could not open VCF file " << params.vcf_file << endl;
exit(1);
}
} else {
cerr << "realignment requires VCF file" << endl;
exit(1);
}
vcf::Variant var(vcffile);
BamAlignment alignment;
map<long int, vector<BamAlignment> > alignmentSortQueue;
// get alignment
// assemble DAG in region around alignment
// loop for each alignment in BAM:
// update DAG when current alignment gets close to edge of assembled DAG
// attempt to realign if read has a certain number of mismatches + gaps or softclips, weighted by basequal
// if alignment to DAG has fewer mismatches and gaps than original alignment, use it
// flatten read into reference space (for now just output alleles from VCF un-spanned insertions)
// write read to queue for streaming re-sorting (some positional change will occur)
long int dag_start_position = 0;
string currentSeqname;
string ref;
//vector<Cigar> cigars; // contains the Cigar strings of nodes in the graph
//vector<long int> refpositions; // contains the reference start coords of nodes in the graph
ReferenceMappings ref_map;
gssw_graph* graph = gssw_graph_create(0);
int8_t* nt_table = gssw_create_nt_table();
int8_t* mat = gssw_create_score_matrix(params.match, params.mism);
int total_reads = 0;
int total_realigned = 0;
int total_improved = 0;
bool emptyDAG = false; // if the dag is constructed over empty sequence
// such as when realigning reads mapped to all-N sequence
if (params.debug) {
cerr << "about to start processing alignments" << endl;
}
while (reader.GetNextAlignment(alignment)) {
string& seqname = referenceIDToName[alignment.RefID];
if (params.debug) {
cerr << "--------------------------------------------" << endl
<< "processing alignment " << alignment.Name << " at "
<< seqname << ":" << alignment.Position << endl;
}
/*
if (!alignment.IsMapped() && graph->size == 0) {
if (params.debug) {
cerr << "unable to build DAG using unmapped read "
<< alignment.Name << " @ "
<< seqname << ":" << alignment.Position
<< " no previous mapped read found and DAG currently empty" << endl;
}
alignmentSortQueue[dag_start_position+dag_window_size].push_back(alignment);
continue;
}
*/
++total_reads;
BamAlignment originalAlignment = alignment;
long unsigned int initialAlignmentPosition = alignment.Position;
//if (dag_start_position == 1) {
// dag_start_position = max(1, (int)initialAlignmentPosition - dag_window_size/2);
//}
// should we construct a new DAG? do so when 3/4 of the way through the current one
// center on current position + 1/2 dag window
// TODO check this scheme using some scribbles on paper
// alignment.IsMapped()
if ((seqname != currentSeqname
|| ((alignment.Position + (alignment.QueryBases.size()/2)
> (3*dag_window_size/4) + dag_start_position)))
&& alignment.Position < reference.sequenceLength(seqname)) {
if (seqname != currentSeqname) {
if (params.debug) {
cerr << "switched ref seqs" << endl;
}
dag_start_position = max((long int) 0,
(long int) (alignment.GetEndPosition() - dag_window_size/2));
// recenter DAG
} else if (!ref_map.empty()) {
dag_start_position = dag_start_position + dag_window_size/2;
dag_start_position = max(dag_start_position,
(long int) (alignment.GetEndPosition() - dag_window_size/2));
} else {
dag_start_position = alignment.Position - dag_window_size/2;
}
dag_start_position = max((long int)0, dag_start_position);
// TODO get sequence length and use to bound noted window size (edge case)
//cerr << "getting ref " << seqname << " " << max((long int) 0, dag_start_position) << " " << dag_window_size << endl;
// get variants for new DAG
vector<vcf::Variant> variants;
if (!vcffile.setRegion(seqname,
dag_start_position + 1,
dag_start_position + dag_window_size)) {
// this is not necessarily an error; there should be a better way to check for VCF file validity
/*
cerr << "could not set region on VCF file to " << currentSeqname << ":"
<< dag_start_position << "-" << dag_start_position + ref.size()
<< endl;
*/
//exit(1);
} else {
// check first variant
if (vcffile.getNextVariant(var)) {
while (var.position <= dag_start_position + 1) {
//cerr << "var position == dag_start_position " << endl;
dag_start_position -= 1;
vcffile.setRegion(seqname,
dag_start_position + 1,
dag_start_position + dag_window_size);
if (!vcffile.getNextVariant(var)) { break; }
}
}
vcffile.setRegion(seqname,
dag_start_position + 1,
dag_start_position + dag_window_size);
while (vcffile.getNextVariant(var)) {
if (params.debug) cerr << "getting variant at " << var.sequenceName << ":" << var.position << endl;
//cerr << var.position << " + " << var.ref.length() << " <= " << dag_start_position << " + " << dag_window_size << endl;
//cerr << var.position << " >= " << dag_start_position << endl;
if (var.position + var.ref.length() <= dag_start_position + dag_window_size
&& var.position >= dag_start_position) {
variants.push_back(var);
}
}
}
//cerr << "dag_start_position " << dag_start_position << endl;
ref = reference.getSubSequence(seqname,
max((long int) 0, dag_start_position),
dag_window_size); // 0/1 conversion
// clear graph and metadata
ref_map.clear();
//cigars.clear();
//refpositions.clear();
gssw_graph_destroy(graph);
if (params.debug) { cerr << "constructing DAG" << endl; }
// and build the DAG
graph = gssw_graph_create(0);
constructDAGProgressive(graph,
ref_map,
ref,
seqname,
variants,
dag_start_position,
nt_table,
mat,
params.flat_input_vcf);
if (params.debug) {
cerr << "graph has " << graph->size << " nodes" << endl;
cerr << "DAG generated from input variants over "
<< seqname << ":" << dag_start_position << "-" << dag_start_position + dag_window_size
<< endl;
}
if (params.display_dag) {
gssw_graph_print(graph);
/*
for (Backbone::iterator b = backbone.begin(); b != backbone.end(); ++b) {
cout << b->first << " "
<< b->first->id << " "
<< b->second.ref_position << " "
<< b->second.cigar << endl
<< b->first->seq << endl;
}
*/
}
if (graph->size == 1 && allN(ref) || graph->size == 0) {
if (params.debug) {
cerr << "DAG is empty (1 node, all N). Alignment is irrelevant." << endl;
}
emptyDAG = true;
} else {
emptyDAG = false;
}
}
AlignmentStats stats_before;
bool was_mapped = alignment.IsMapped();
bool has_realigned = false;
if (was_mapped) {
if (dag_start_position + dag_window_size < alignment.GetEndPosition()) {
ref = reference.getSubSequence(seqname,
max((long int) 0, dag_start_position),
alignment.GetEndPosition() - dag_start_position); // 0/1 conversion
}
}
if (params.debug) {
if (emptyDAG) {
cerr << "cannot realign against empty (all-N single node) graph" << endl;
}
}
if (!emptyDAG && shouldRealign(alignment, ref, dag_start_position, params, stats_before)) {
++total_realigned;
if (params.debug) {
cerr << "realigning: " << alignment.Name
<< " " << alignment.QueryBases << endl
<< " aligned @ " << alignment.Position
<< " to variant graph over "
<< seqname
<< ":" << dag_start_position
<< "-" << dag_start_position + dag_window_size << endl;
}
//{
try {
Cigar flat_cigar;
string read = alignment.QueryBases;
string qualities = alignment.Qualities;
int score;
long int position;
string strand;
gssw_graph_mapping* gm =
gswalign(graph,
ref_map,
read,
qualities,
params,
position,
score,
flat_cigar,
strand,
nt_table,
mat);
//
gssw_graph_mapping_destroy(gm);
if (params.dry_run) {
if (strand == "-" && !alignment.IsMapped()) {
read = reverseComplement(read);
}
cout << read << endl;
cout << graph_mapping_to_string(gm) << endl;
cout << score << " " << strand << " "
<< position << " "
<< flat_cigar << endl;
} else {
/*
if (strand == "-") {
read = reverseComplement(trace_report.read);
}
*/
// TODO the qualities are not on the right side of the read
if (strand == "-" && alignment.IsMapped()) {
// if we're realigning, this is always true unless we swapped strands
alignment.SetIsReverseStrand(true);
//reverse(alignment.Qualities.begin(), alignment.Qualities.end()); // reverse qualities
}
//alignment.QueryBases = reverseComplement(trace_report.read);
alignment.QueryBases = read;
alignment.Qualities = qualities;
alignment.Position = position;// + 1;// + 1;//(trace_report.node->position - 1) + trace_report.x;
alignment.SetIsMapped(true);
if (!alignment.MapQuality) {
alignment.MapQuality = 20; // horrible hack... at least approximate with alignment mismatches against graph
}
// check if somehow we've ended up with an indel at the ends
// if so, grab the reference sequence right beyond it and add
// a single match to the cigar, allowing variant detection methods
// to run on the results without internal modification
Cigar& cigar = flat_cigar;
//cerr << flat_cigar << " " << flat_cigar.readLen() << " " << flat_cigar.refLen() << endl;
int flankSize = params.flatten_flank;
if (cigar.front().isIndel() ||
(cigar.front().isSoftclip() && cigar.at(1).isIndel())) {
alignment.Position -= flankSize;
string refBase = reference.getSubSequence(seqname, alignment.Position, flankSize);
if (cigar.front().isSoftclip()) {
alignment.QueryBases.erase(alignment.QueryBases.begin(),
alignment.QueryBases.begin()+cigar.front().length);
alignment.Qualities.erase(alignment.Qualities.begin(),
alignment.Qualities.begin()+cigar.front().length);
cigar.erase(cigar.begin());
}
alignment.QueryBases.insert(0, refBase);
alignment.Qualities.insert(0, string(flankSize, shortInt2QualityChar(30)));
Cigar newCigar; newCigar.push_back(CigarElement(flankSize, 'M'));
newCigar.append(flat_cigar);
flat_cigar = newCigar;
}
if (cigar.back().isIndel() ||
(cigar.back().isSoftclip() && cigar.at(cigar.size()-2).isIndel())) {
string refBase = reference.getSubSequence(seqname,
alignment.Position
+ flat_cigar.refLen(),
flankSize);
if (cigar.back().isSoftclip()) {
alignment.QueryBases.erase(alignment.QueryBases.end()-cigar.back().length,
alignment.QueryBases.end());
alignment.Qualities.erase(alignment.Qualities.end()-cigar.back().length,
alignment.Qualities.end());
cigar.pop_back();
}
Cigar newCigar; newCigar.push_back(CigarElement(flankSize, 'M'));
flat_cigar.append(newCigar);
//flat_cigar.append(newCigar);
alignment.QueryBases.append(refBase);
alignment.Qualities.append(string(flankSize, shortInt2QualityChar(30)));
}
flat_cigar.toCigarData(alignment.CigarData);
//cerr << flat_cigar << " " << flat_cigar.readLen() << " " << flat_cigar.refLen() << endl;
if (dag_start_position + dag_window_size < alignment.GetEndPosition()) {
ref = reference.getSubSequence(seqname,
max((long int) 0, dag_start_position),
alignment.GetEndPosition() - dag_start_position); // 0/1 conversion
}
AlignmentStats stats_after;
countMismatchesAndGaps(alignment, flat_cigar, ref, dag_start_position, stats_after, params.debug);
/*
if ((!was_mapped || (stats_before.softclip_qsum >= stats_after.softclip_qsum
&& stats_before.mismatch_qsum >= stats_after.mismatch_qsum))
&& acceptRealignment(alignment, ref, dag_start_position, params, stats_after)) {
*/
/*
if ((!was_mapped || (stats_before.softclip_qsum + stats_before.mismatch_qsum
>= stats_after.softclip_qsum + stats_after.mismatch_qsum))
&& acceptRealignment(alignment, ref, dag_start_position, params, stats_after)) {
*/
// we accept the new alignment if...
if (!was_mapped // it wasn't mapped previously
// or if we have removed soft clips or mismatches (per quality) from the alignment
//|| ((stats_before.softclip_qsum >= stats_after.softclip_qsum
// && stats_before.mismatch_qsum >= stats_after.mismatch_qsum)
|| ((stats_before.softclip_qsum + stats_before.mismatch_qsum
>= stats_after.softclip_qsum + stats_after.mismatch_qsum)
// and if we have added gaps, we have added them to remove mismatches or softclips
&& (stats_before.gaps >= stats_after.gaps // accept any time we reduce gaps while not increasing softclips/mismatches
|| (stats_before.gaps < stats_after.gaps // and allow gap increases when they improve the alignment
&& (stats_before.softclip_qsum
+ stats_before.mismatch_qsum
>
stats_after.softclip_qsum
+ stats_after.mismatch_qsum))))
// and the alignment must not have more than the acceptable number of gaps, softclips, or mismatches
// as provided in input parameters
&& acceptRealignment(alignment, ref, dag_start_position, params, stats_after)) {
// keep the alignment
// TODO require threshold of softclips to keep alignment (or count of gaps, mismatches,...)
if (params.debug) {
cerr << "realigned " << alignment.Name << " to graph, which it maps to with "
<< stats_after.mismatch_qsum << "q in mismatches and "
<< stats_after.softclip_qsum << "q in soft clips" << endl;
}
++total_improved;
has_realigned = true;
} else {
// reset to old version of alignment
if (params.debug) {
cerr << "failed realignment of " << alignment.Name << " to graph, which it maps to with: "
<< stats_after.mismatch_qsum << "q in mismatches " << "(vs " << stats_before.mismatch_qsum << "q before), and "
<< stats_after.softclip_qsum << "q in soft clips " << "(vs " << stats_before.softclip_qsum << "q before) " << endl;
}
has_realigned = false;
alignment = originalAlignment;
}
}
//} // try block
} catch (...) {
cerr << "exception when realigning " << alignment.Name
<< " at position " << referenceIDToName[alignment.RefID]
<< ":" << alignment.Position
<< " " << alignment.QueryBases << endl;
// reset to original alignment
has_realigned = false;
alignment = originalAlignment;
}
}
// ensure correct order if alignments move
long int maxOutputPos = initialAlignmentPosition - dag_window_size;
// if we switched sequences we need to flush out all the reads from the previous one
string lastSeqname = currentSeqname;
if (seqname != currentSeqname) {
// so the max output position is set past the end of the last chromosome
if (!currentSeqname.empty()) {
maxOutputPos = reference.sequenceLength(currentSeqname) + dag_window_size;
}
currentSeqname = seqname;
}
if (!params.dry_run) {
map<long int, vector<BamAlignment> >::iterator p = alignmentSortQueue.begin();
for ( ; p != alignmentSortQueue.end(); ++p) {
// except if we are running in unsorted mode, stop when we are at the window size
if (!params.unsorted_output && p->first > maxOutputPos) {
break; // no more to do
} else {
for (vector<BamAlignment>::iterator a = p->second.begin(); a != p->second.end(); ++a) {
writer.SaveAlignment(*a);
}
}
}
if (p != alignmentSortQueue.begin()) {
alignmentSortQueue.erase(alignmentSortQueue.begin(), p);
}
if (!params.only_realigned || has_realigned) {
alignmentSortQueue[alignment.Position].push_back(alignment);
}
}
} // end GetNextAlignment loop
if (!params.dry_run) {
map<long int, vector<BamAlignment> >::iterator p = alignmentSortQueue.begin();
for ( ; p != alignmentSortQueue.end(); ++p) {
for (vector<BamAlignment>::iterator a = p->second.begin(); a != p->second.end(); ++a)
writer.SaveAlignment(*a);
}
}
gssw_graph_destroy(graph);
free(nt_table);
free(mat);
reader.Close();
writer.Close();
if (params.debug) {
cerr << "total reads:\t" << total_reads << endl;
cerr << "realigned:\t" << total_realigned << endl;
cerr << "improved:\t" << total_improved << endl;
}
}
