bool BamProcessor::processRecord ()
{
trclog << "\nProcessing record " << read_cnt_ << " - " << rec_.getReadName () << ", " << rec_.get0BasedUnclippedEnd () << "->" << rec_.getReadLength () << ", ref " << rec_.getReferenceName () << std::endl;
const char* seq = rec_.getSequence ();
unsigned position = rec_.get0BasedPosition ();
unsigned new_position = position;
bool reverse_match = (rec_.getFlag () & 0x10);
Cigar* cigar_p = rec_.getCigarInfo ();
if (!cigar_p->size ()) // can not recreate reference is cigar is missing. Keep record unaligned.
{ // TODO: allow to specify and load external reference
++ unaligned_cnt_;
return true;
}
myassert (cigar_p);
const String *mdval = rec_.getStringTag ("MD");
if (!mdval) // can not recreate reference is MD tag is missing. Keep record as is.
{
warn << "No MD Tag for record " << proc_cnt_ << ". Skipping record." << std::endl;
++nomd_cnt_;
return true; // record will be kept as-is.
}
std::string md_tag = mdval->c_str ();
// find the non-clipped region
uint32_t clean_len;
EndClips clips;
const char* clean_read = clip_seq (seq, *cigar_p, clean_len, clips);
// find length needed for the reference
// this reserves space enough for entire refference, including softclipped ends.
unsigned ref_len = cigar_p->getExpectedReferenceBaseCount ();
if (ref_buffer_sz_ < ref_len)
{
ref_buffer_sz_ = (1 + ref_len / REF_BUF_INCR) * REF_BUF_INCR;
ref_buffer_.reset (ref_buffer_sz_);
}
if (clean_len > MAX_SEQ_LEN || ref_len > MAX_SEQ_LEN)
{
++ toolongs_;
return true;
}
// recreate reference by Query, Cigar, and MD tag. Do not include softclipped ends in the recreated sequence (use default last parameter)
recreate_ref (seq, rec_.getReadLength (), cigar_p, md_tag.c_str (), ref_buffer_, ref_buffer_sz_);
unsigned qry_ins; // extra bases in query == width_left
unsigned ref_ins; // extra bases in reference == width_right
band_width (*cigar_p, qry_ins, ref_ins);
if (log_matr_ || log_base_)
{
logfile_ << "Record " << read_cnt_ << ": " << rec_.getReadName () << "\n"
<< " sequence (" << rec_.getReadLength () << " bases)\n";
}
CigarRoller roller;
int ref_shift = 0; // shift of the new alignment position on refereance relative the original
unsigned qry_off, ref_off; // offsets on the query and reference of the first non-clipped aligned bases
double new_score = 0;
switch (p_->algo ())
{
case ContalignParams::TEMPL:
{
// call aligner
new_score = taligner_.eval (clean_read, clean_len, ref_buffer_, ref_len, 0, band_width_);
// read traceback
// TODO: convert directly to cigar
genstr::Alignment* al = taligner_.trace ();
// convert alignment to cigar
ref_shift = roll_cigar (roller, *al, clean_len, clips, qry_off, ref_off);
}
break;
case ContalignParams::PLAIN:
{
new_score = aligner_.align_band (
clean_read, // xseq
clean_len, // xlen
ref_buffer_, // yseq
ref_len, // ylen
0, // xpos
0, // ypos
std::max (clean_len, ref_len), // segment length
qry_ins + band_width_, // width_left
false, // unpack
ref_ins + band_width_, // width_right - forces to width_left
true, // to_beg
true // to_end
);
unsigned bno = aligner_.backtrace (
batches_, // BATCH buffer
max_batch_no_, // size of BATCH buffer
false, // fill the BATCH array in reverse direction
ref_ins + band_width_ // width
);
// convert alignment to cigar
ref_shift = roll_cigar (roller, batches_, bno, clean_len, clips, qry_off, ref_off);
}
break;
case ContalignParams::POLY:
{
new_score = contalign_.align_band (
clean_read, // xseq
clean_len, // xlen
ref_buffer_, // yseq
ref_len, // ylen
0, // xpos
0, // ypos
std::max (clean_len, ref_len), // segment length
qry_ins + band_width_, // width_left
false, // unpack
ref_ins + band_width_, // width_right - forces to width_left
true, // to_beg
true // to_end
);
unsigned bno = contalign_.backtrace (
batches_, // BATCH buffer
max_batch_no_, // size of BATCH buffer
false, // fill the BATCH array in reverse direction
ref_ins + band_width_ // width
);
// convert alignment to cigar
ref_shift = roll_cigar (roller, batches_, bno, clean_len, clips, qry_off, ref_off);
}
break;
default:
break;
}
++realigned_cnt_;
// compare original and new cigar (and location)
if (ref_shift || !(*cigar_p == roller))
{
// save original cigar and position for reporting
std::string orig_cigar_str;
rec_.getCigarInfo ()->getCigarString (orig_cigar_str);
int32_t prior_pos = rec_.get0BasedPosition ();
// replace cigar
rec_.setCigar (roller);
++ modified_cnt_;
// update pos_adjusted_cnt if position changed
if (ref_shift != 0)
{
myassert (prior_pos + ref_shift >= 0);
rec_.set0BasedPosition (prior_pos + ref_shift);
++ pos_adjusted_cnt_;
}
if (log_diff_)
{
const unsigned MAX_BATCH_PRINTED = 100;
BATCH batches [MAX_BATCH_PRINTED];
std::string new_cigar_str;
unsigned bno;
int swscore;
rec_.getCigarInfo ()->getCigarString (new_cigar_str);
if (!log_base_ && !log_matr_)
logfile_ << "Record " << read_cnt_ << ": " << rec_.getReadName () << " (" << rec_.getReadLength () << " bases)\n";
logfile_ << " ORIG ALIGNMENT:" << std::right << std::setw (9) << prior_pos+1 << "->" << orig_cigar_str << "\n";
bno = cigar_to_batches (orig_cigar_str, batches, MAX_BATCH_PRINTED);
swscore = align_score (batches, bno, clean_read, ref_buffer_, p_->gip (), p_->gep (), p_->mat (), p_->mis ());
print_batches (clean_read, clean_len, false, ref_buffer_, ref_len, false, batches, bno, logfile_, false, prior_pos + clips.soft_beg_, clips.soft_beg_, 0, 160);
logfile_ << "\n 'classic' SW score is " << swscore << "\n";
logfile_ << " NEW ALIGNMENT:" << std::right << std::setw (9) << rec_.get1BasedPosition () << "->" << new_cigar_str << std::endl;
bno = cigar_to_batches (new_cigar_str, batches, MAX_BATCH_PRINTED);
swscore = align_score (batches, bno, clean_read + qry_off, ref_buffer_ + ref_off, p_->gip (), p_->gep (), p_->mat (), p_->mis ());
print_batches (clean_read + qry_off, clean_len - qry_off, false, ref_buffer_ + ref_off, ref_len - ref_off, false, batches, bno, logfile_, false, prior_pos + clips.soft_beg_ + ref_off, clips.soft_beg_ + qry_off, 0, 160);
logfile_ << "\n 'classic' SW score is " << swscore;
logfile_ << "\n alternate (context-aware) score is " << new_score << ", used bandwidth left: " << qry_ins + band_width_ << ", right: " << ref_ins + band_width_ << "\n" << std::endl;
}
else if (log_base_)
{
logfile_ << "Recomputed alignment differs from original:\n";
logfile_ << " ORIG ALIGNMENT:" << std::right << std::setw (9) << prior_pos+1 << "->" << orig_cigar_str << "\n";
std::string new_cigar_str;
rec_.getCigarInfo ()->getCigarString (new_cigar_str);
logfile_ << " NEW ALIGNMENT:" << std::right << std::setw (9) << rec_.get1BasedPosition () << "->" << new_cigar_str << "\n" << std::endl;
}
}
else
{
if (log_base_)
{
logfile_ << "Recomputed alignment matches the original:\n";
std::string orig_cigar_str;
rec_.getCigarInfo ()->getCigarString (orig_cigar_str);
int32_t prior_pos = rec_.get0BasedPosition ();
logfile_ << " " << std::right << std::setw (9) << prior_pos+1 << "->" << orig_cigar_str << "\n" << std::endl;
}
}
return true;
}
void validateRead1ModQuality(SamRecord& samRecord)
{
//////////////////////////////////////////
// Validate Record 1
// Create record structure for validating.
int expectedBlockSize = 89;
const char* expectedReferenceName = "1";
const char* expectedMateReferenceName = "1";
const char* expectedMateReferenceNameOrEqual = "=";
bamRecordStruct* expectedRecordPtr =
(bamRecordStruct *) malloc(expectedBlockSize + sizeof(int));
char tag[3];
char type;
void* value;
bamRecordStruct* bufferPtr;
unsigned char* varPtr;
expectedRecordPtr->myBlockSize = expectedBlockSize;
expectedRecordPtr->myReferenceID = 0;
expectedRecordPtr->myPosition = 1010;
expectedRecordPtr->myReadNameLength = 23;
expectedRecordPtr->myMapQuality = 0;
expectedRecordPtr->myBin = 4681;
expectedRecordPtr->myCigarLength = 2;
expectedRecordPtr->myFlag = 73;
expectedRecordPtr->myReadLength = 5;
expectedRecordPtr->myMateReferenceID = 0;
expectedRecordPtr->myMatePosition = 1010;
expectedRecordPtr->myInsertSize = 0;
// Check the alignment end
assert(samRecord.get0BasedAlignmentEnd() == 1016);
assert(samRecord.get1BasedAlignmentEnd() == 1017);
assert(samRecord.getAlignmentLength() == 7);
assert(samRecord.get0BasedUnclippedStart() == 1010);
assert(samRecord.get1BasedUnclippedStart() == 1011);
assert(samRecord.get0BasedUnclippedEnd() == 1016);
assert(samRecord.get1BasedUnclippedEnd() == 1017);
// Check the accessors.
assert(samRecord.getBlockSize() == expectedRecordPtr->myBlockSize);
assert(samRecord.getReferenceID() == expectedRecordPtr->myReferenceID);
assert(strcmp(samRecord.getReferenceName(), expectedReferenceName) == 0);
assert(samRecord.get1BasedPosition() == expectedRecordPtr->myPosition + 1);
assert(samRecord.get0BasedPosition() == expectedRecordPtr->myPosition);
assert(samRecord.getReadNameLength() ==
expectedRecordPtr->myReadNameLength);
assert(samRecord.getMapQuality() == expectedRecordPtr->myMapQuality);
assert(samRecord.getBin() == expectedRecordPtr->myBin);
assert(samRecord.getCigarLength() == expectedRecordPtr->myCigarLength);
assert(samRecord.getFlag() == expectedRecordPtr->myFlag);
assert(samRecord.getReadLength() == expectedRecordPtr->myReadLength);
assert(samRecord.getMateReferenceID() ==
expectedRecordPtr->myMateReferenceID);
assert(strcmp(samRecord.getMateReferenceName(),
expectedMateReferenceName) == 0);
assert(strcmp(samRecord.getMateReferenceNameOrEqual(),
expectedMateReferenceNameOrEqual) == 0);
assert(samRecord.get1BasedMatePosition() ==
expectedRecordPtr->myMatePosition + 1);
assert(samRecord.get0BasedMatePosition() ==
expectedRecordPtr->myMatePosition);
assert(samRecord.getInsertSize() == expectedRecordPtr->myInsertSize);
assert(strcmp(samRecord.getReadName(), "1:1011:F:255+17M15D20M") == 0);
assert(strcmp(samRecord.getCigar(), "5M2D") == 0);
assert(strcmp(samRecord.getSequence(), "CCGAA") == 0);
assert(strcmp(samRecord.getQuality(), "ABCDE") == 0);
assert(samRecord.getNumOverlaps(1010, 1017) == 5);
assert(samRecord.getNumOverlaps(1010, 1016) == 5);
assert(samRecord.getNumOverlaps(1012, 1017) == 3);
assert(samRecord.getNumOverlaps(1015, 1017) == 0);
assert(samRecord.getNumOverlaps(1017, 1010) == 0);
assert(samRecord.getNumOverlaps(1013, 1011) == 0);
assert(samRecord.getNumOverlaps(-1, 1017) == 5);
// Reset the tag iter, since the tags have already been read.
samRecord.resetTagIter();
// Check the tags.
assert(samRecord.getNextSamTag(tag, type, &value) == true);
assert(tag[0] == 'A');
assert(tag[1] == 'M');
assert(type == 'i');
assert(*(char*)value == 0);
assert(samRecord.getNextSamTag(tag, type, &value) == true);
assert(tag[0] == 'M');
assert(tag[1] == 'D');
assert(type == 'Z');
assert(*(String*)value == "37");
assert(samRecord.getNextSamTag(tag, type, &value) == true);
assert(tag[0] == 'N');
assert(tag[1] == 'M');
assert(type == 'i');
assert(*(char*)value == 0);
assert(samRecord.getNextSamTag(tag, type, &value) == true);
assert(tag[0] == 'X');
assert(tag[1] == 'T');
assert(type == 'A');
assert(*(char*)value == 'R');
// No more tags, should return false.
assert(samRecord.getNextSamTag(tag, type, &value) == false);
assert(samRecord.getNextSamTag(tag, type, &value) == false);
// Get the record ptr.
bufferPtr = (bamRecordStruct*)samRecord.getRecordBuffer();
// Validate the buffers match.
assert(bufferPtr->myBlockSize == expectedRecordPtr->myBlockSize);
assert(bufferPtr->myReferenceID == expectedRecordPtr->myReferenceID);
assert(bufferPtr->myPosition == expectedRecordPtr->myPosition);
assert(bufferPtr->myReadNameLength == expectedRecordPtr->myReadNameLength);
assert(bufferPtr->myMapQuality == expectedRecordPtr->myMapQuality);
assert(bufferPtr->myBin == expectedRecordPtr->myBin);
assert(bufferPtr->myCigarLength == expectedRecordPtr->myCigarLength);
assert(bufferPtr->myFlag == expectedRecordPtr->myFlag);
assert(bufferPtr->myReadLength == expectedRecordPtr->myReadLength);
assert(bufferPtr->myMateReferenceID ==
expectedRecordPtr->myMateReferenceID);
assert(bufferPtr->myMatePosition == expectedRecordPtr->myMatePosition);
assert(bufferPtr->myInsertSize == expectedRecordPtr->myInsertSize);
// Validate the variable length fields in the buffer.
// Set the pointer to the start of the variable fields.
varPtr = (unsigned char*)(&(bufferPtr->myData[0]));
// Validate the readname.
for(int i = 0; i < expectedRecordPtr->myReadNameLength; i++)
{
assert(*varPtr == samRecord.getReadName()[i]);
varPtr++;
}
// Validate the cigar.
// The First cigar is 5M which is 5 << 4 | 0 = 80
assert(*(unsigned int*)varPtr == 80);
// Increment the varptr the size of an int.
varPtr += 4;
// The 2nd cigar is 2D which is 2 << 4 | 2 = 34
assert(*(unsigned int*)varPtr == 34);
// Increment the varptr the size of an int.
varPtr += 4;
// Validate the sequence.
// CC = 0x22
assert(*varPtr == 0x22);
varPtr++;
// GA = 0x41
assert(*varPtr == 0x41);
varPtr++;
// A = 0x10
assert(*varPtr == 0x10);
varPtr++;
// Validate the Quality
for(int i = 0; i < expectedRecordPtr->myReadLength; i++)
{
assert(*varPtr == samRecord.getQuality()[i] - 33);
varPtr++;
}
// Validate the tags.
assert(*varPtr == 'A');
varPtr++;
assert(*varPtr == 'M');
varPtr++;
assert(*varPtr == 'C');
varPtr++;
assert(*varPtr == 0);
varPtr++;
assert(*varPtr == 'M');
varPtr++;
assert(*varPtr == 'D');
varPtr++;
assert(*varPtr == 'Z');
varPtr++;
assert(*varPtr == '3');
varPtr++;
assert(*varPtr == '7');
varPtr++;
assert(*varPtr == 0);
varPtr++;
assert(*varPtr == 'N');
varPtr++;
assert(*varPtr == 'M');
varPtr++;
assert(*varPtr == 'C');
varPtr++;
assert(*varPtr == 0);
varPtr++;
assert(*varPtr == 'X');
varPtr++;
assert(*varPtr == 'T');
varPtr++;
assert(*varPtr == 'A');
varPtr++;
assert(*varPtr == 'R');
varPtr++;
}
