// prepare internal structures for clipping and alignment
// returns true if realignment was performed
bool ContAlignImp::compute_alignment (
const char* q_seq,
unsigned q_len,
const char* r_seq,
unsigned r_len,
int r_pos,
bool forward,
const uint32_t* cigar,
unsigned cigar_sz,
uint32_t*& cigar_dest,
unsigned& cigar_dest_sz,
int& new_pos,
bool& already_perfect,
bool& clip_failed,
bool& alignment_failed,
bool& unclip_failed)
{
already_perfect = false;
alignment_failed = false;
unclip_failed = false;
// unsigned oplen;
//const char* q_seq_clipped = q_seq;
//const uint32_t* cigar_clipped = cigar;
//unsigned cigar_sz_clipped = cigar_sz;
//unsigned sclip_q_len, sclip_r_len, sclip_al_len;
assert (cigar_sz);
// clip out the hard and soft clipping zones from 5" and 3"
// The 'cut out' of the q_seq is done by switching to downstream pointer.
uint32_t clean_len;
EndClips clips;
const char* clean_read = clip_seq (q_seq, cigar, cigar_sz, clean_len, clips);
// clip reference accordingly
//r_seq += clips.soft_beg_;
//r_len -= clips.soft_beg_ + clips.soft_end_;
if (clean_len > MAX_SEQ_LEN || r_len > MAX_SEQ_LEN)
{
// std::cerr << "Sequence is too long to fit into aligner (" << std::max (clean_len, r_len) << ", max is " << MAX_SEQ_LEN << ")" << std::endl;
#if 0
std::cerr << "Cigar is ";
cigar_out (std::cerr, cigar, cigar_sz);
std::cerr << "\nClips: " << clips << std::endl;
#endif
return false;
}
unsigned qry_ins; // extra bases in query == width_left
unsigned ref_ins; // extra bases in reference == width_right
band_width (cigar, cigar_sz, qry_ins, ref_ins);
if (!ContAlign::can_align (clean_len, r_len, qry_ins + extra_bandwidth_, ref_ins + extra_bandwidth_))
return false;
ContAlign::align_band (
clean_read, // xseq
clean_len, // xlen
r_seq, // yseq
r_len, // ylen
0, // xpos
0, // ypos
std::max (clean_len, r_len), // segment length
qry_ins + extra_bandwidth_, // width_left
ref_ins + extra_bandwidth_, // 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
ref_ins + extra_bandwidth_ // width
);
// convert alignment to cigar
unsigned qry_off, ref_off;
// int ref_shift =
roll_cigar (new_cigar_, MAX_CIGAR_SZ, cigar_dest_sz, batches_, bno, clean_len, clips, qry_off, ref_off);
new_pos = r_pos + ref_off;
cigar_dest = new_cigar_;
return true;
}
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;
}