/**
* @brief Debug dump, optional
*/
void KlibAlignment::update()
{
_impl->result = ksw_align(
_impl->reflen, _impl->ref.get(),
_impl->altlen, _impl->alt.get(),
5, _impl->mat, _impl->gapo, _impl->gape,
KSW_XSTART, // add flags here
&(_impl->qprofile));
if(_impl->cigar)
{
free(_impl->cigar);
_impl->cigar = NULL;
_impl->cigar_len = 0;
}
ksw_global(
_impl->result.qe - _impl->result.qb + 1,
_impl->ref.get() + _impl->result.qb,
_impl->result.te - _impl->result.tb + 1,
_impl->alt.get() + _impl->result.tb,
5, _impl->mat, _impl->gapo, _impl->gape,
_impl->altlen,
&_impl->cigar_len, &_impl->cigar);
_impl->valid_result = true;
}
void bsw2_pair1(const bsw2opt_t *opt, int64_t l_pac, const uint8_t *pac, const bsw2pestat_t *st, const bsw2hit_t *h, int l_mseq, const char *mseq, bsw2hit_t *a, int8_t g_mat[25])
{
extern void seq_reverse(int len, ubyte_t *seq, int is_comp);
int64_t k, beg, end;
uint8_t *seq, *ref;
int i;
// compute the region start and end
a->n_seeds = 1; a->flag |= BSW2_FLAG_MATESW; // before calling this routine, *a has been cleared with memset(0); the flag is set with 1<<6/7
if (h->is_rev == 0) {
beg = (int64_t)(h->k + st->avg - EXT_STDDEV * st->std - l_mseq + .499);
if (beg < h->k) beg = h->k;
end = (int64_t)(h->k + st->avg + EXT_STDDEV * st->std + .499);
a->is_rev = 1; a->flag |= 16;
} else {
beg = (int64_t)(h->k + h->end - h->beg - st->avg - EXT_STDDEV * st->std + .499);
end = (int64_t)(h->k + h->end - h->beg - st->avg + EXT_STDDEV * st->std + l_mseq + .499);
if (end > h->k + (h->end - h->beg)) end = h->k + (h->end - h->beg);
a->is_rev = 0;
}
if (beg < 1) beg = 1;
if (end > l_pac) end = l_pac;
if (end - beg < l_mseq) return;
// generate the sequence
seq = malloc(l_mseq + (end - beg));
ref = seq + l_mseq;
for (k = beg; k < end; ++k)
ref[k - beg] = pac[k>>2] >> ((~k&3)<<1) & 0x3;
if (h->is_rev == 0) {
for (i = 0; i < l_mseq; ++i) { // on the reverse strand
int c = nst_nt4_table[(int)mseq[i]];
seq[l_mseq - 1 - i] = c > 3? 4 : 3 - c;
}
} else {
for (i = 0; i < l_mseq; ++i) // on the forward strand
seq[i] = nst_nt4_table[(int)mseq[i]];
}
{
int flag = KSW_XSUBO | KSW_XSTART | (l_mseq * g_mat[0] < 250? KSW_XBYTE : 0) | opt->t;
kswr_t aln;
aln = ksw_align(l_mseq, seq, end - beg, ref, 5, g_mat, opt->q, opt->r, flag, 0);
a->G = aln.score;
a->G2 = aln.score2;
if (a->G < opt->t) a->G = 0;
if (a->G2 < opt->t) a->G2 = 0;
if (a->G2) a->flag |= BSW2_FLAG_TANDEM;
a->k = beg + aln.tb;
a->len = aln.te - aln.tb + 1;
a->beg = aln.qb;
a->end = aln.qe + 1;
/*
printf("[Q] "); for (i = 0; i < l_mseq; ++i) putchar("ACGTN"[(int)seq[i]]); putchar('\n');
printf("[R] "); for (i = 0; i < end - beg; ++i) putchar("ACGTN"[(int)ref[i]]); putchar('\n');
printf("G=%d,G2=%d,beg=%d,end=%d,k=%lld,len=%d\n", a->G, a->G2, a->beg, a->end, a->k, a->len);
*/
}
if (a->is_rev) i = a->beg, a->beg = l_mseq - a->end, a->end = l_mseq - i;
free(seq);
}
void ta_trim1(ta_opt_t *opt, char *seq)
{
int i, j, k;
kstring_t _str = {0,0,0}, *str = &_str;
str->l = strlen(seq);
str->s = malloc(str->l);
for (i = 0; i < str->l; ++i)
str->s[i] = seq_nt4_table[(uint8_t)seq[i]];
for (j = 0; j < opt->n_adaps; ++j) {
kswr_t r;
double diff;
int type;
ta_adap_t *p = &opt->adaps[j];
r = ksw_align(p->len, p->seq, str->l, (uint8_t*)str->s, 5, opt->mat, opt->go, opt->ge, KSW_XBYTE|KSW_XSTART|(opt->min_len * opt->sa), 0);
++r.te; ++r.qe; // change to 0-based
k = r.qe - r.qb < r.te - r.tb? r.qe - r.qb : r.te - r.tb;
diff = (double)(k * opt->sa - r.score) / opt->sb / k;
//printf("%d:%.3f [%d,%d):%d <=> [%d,%d):%d\n", r.score, diff, r.qb, r.qe, p->len, r.tb, r.te, (int)str.l);
if (r.qb <= r.tb && p->len - r.qe <= str->l - r.te) { // contained
if (r.qb * opt->sa > opt->sa + opt->sb) continue;
if ((p->len - r.qe) * opt->sa > opt->sa + opt->sb) continue;
type = 1;
} else if (r.qb <= r.tb) { // 3'-end overlap
if (r.qb * opt->sa > opt->sa + opt->sb) continue;
if ((str->l - r.te) * opt->sa > opt->sa + opt->sb) continue;
type = 2;
} else { // 5'-end overlap
if ((p->len - r.qe) * opt->sa > opt->sa + opt->sb) continue;
if (r.tb * opt->sa > opt->sa + opt->sb) continue;
type = 3;
}
if (p->type == 5) {
if (r.tb == 0 && r.qe == p->len && (r.te - r.tb) * opt->sa == r.score)
type = 4;
} else if (p->type == 3) {
if (r.qb == 0 && r.te == str->l && (r.te - r.tb) * opt->sa == r.score)
type = 4;
}
if (type == 4) { // exact match
if (r.te - r.tb < opt->min_len) continue;
} else { // inexact match
if (r.score < opt->min_sc || diff > opt->max_diff) continue;
}
__sync_fetch_and_add(&p->cnt, 1);
if (p->type == 5) {
k = r.te + (p->len - r.qe);
k = k < str->l? k : str->l;
for (i = 0; i < k; ++i) seq[i] = 'X';
} else if (p->type == 3) {
k = r.tb > r.qb? r.tb - r.qb : 0;
for (i = k; i < str->l; ++i) seq[i] = 'X';
}
}
free(str->s);
}
static aln_v align_read(const kseq_t *read,
const kseq_v targets,
const align_config_t *conf)
{
kseq_t *r;
const int32_t read_len = read->seq.l;
aln_v result;
kv_init(result);
kv_resize(aln_t, result, kv_size(targets));
uint8_t *read_num = calloc(read_len, sizeof(uint8_t));
for(size_t k = 0; k < read_len; ++k)
read_num[k] = conf->table[(int)read->seq.s[k]];
// Align to each target
kswq_t *qry = NULL;
for(size_t j = 0; j < kv_size(targets); j++) {
// Encode target
r = &kv_A(targets, j);
uint8_t *ref_num = calloc(r->seq.l, sizeof(uint8_t));
for(size_t k = 0; k < r->seq.l; ++k)
ref_num[k] = conf->table[(int)r->seq.s[k]];
aln_t aln;
aln.target_idx = j;
aln.loc = ksw_align(read_len, read_num,
r->seq.l, ref_num,
conf->m,
conf->mat,
conf->gap_o,
conf->gap_e,
KSW_XSTART,
&qry);
ksw_global(aln.loc.qe - aln.loc.qb + 1,
&read_num[aln.loc.qb],
aln.loc.te - aln.loc.tb + 1,
&ref_num[aln.loc.tb],
conf->m,
conf->mat,
conf->gap_o,
conf->gap_e,
50, /* TODO: Magic number - band width */
&aln.n_cigar,
&aln.cigar);
aln.nm = 0;
size_t qi = aln.loc.qb, ri = aln.loc.tb;
for(size_t k = 0; k < aln.n_cigar; k++) {
const int32_t oplen = bam_cigar_oplen(aln.cigar[k]),
optype = bam_cigar_type(aln.cigar[k]);
if(optype & 3) { // consumes both - check for mismatches
for(size_t j = 0; j < oplen; j++) {
if(UNLIKELY(read_num[qi + j] != ref_num[ri + j]))
aln.nm++;
}
} else {
aln.nm += oplen;
}
if(optype & 1) qi += oplen;
if(optype & 2) ri += oplen;
}
kv_push(aln_t, result, aln);
free(ref_num);
}
free(qry);
free(read_num);
ks_introsort(dec_score, kv_size(result), result.a);
return result;
}
int Aligner::getOffsetAgainstMaster(std::string& slaveDR, AlignerFlag_t& flags) {
int slave_dr_length = static_cast<int>(slaveDR.length());
uint8_t* slave_dr_forward = new uint8_t[slave_dr_length+1];
uint8_t* slave_dr_reverse = new uint8_t[slave_dr_length+1];
prepareSlaveForAlignment(slaveDR, slave_dr_forward, slave_dr_reverse);
// query profile
kswq_t *slave_forward_query_profile = 0;
kswq_t *slave_reverse_query_profile = 0;
// alignment of slave against master
kswr_t forward_return = ksw_align(slave_dr_length,
slave_dr_forward,
AL_masterDRLength,
AL_masterDR,
5,
AL_scoringMatrix,
AL_gapOpening,
AL_gapExtension,
AL_xtra,
&slave_forward_query_profile);
kswr_t reverse_return = ksw_align(slave_dr_length,
slave_dr_reverse,
AL_masterDRLength,
AL_masterDR,
5,
AL_scoringMatrix,
AL_gapOpening,
AL_gapExtension,
AL_xtra,
&slave_reverse_query_profile);
// free the query profile
free(slave_forward_query_profile);
free(slave_reverse_query_profile);
delete slave_dr_reverse;
delete slave_dr_forward;
// figure out which alignment was better
if (reverse_return.score == forward_return.score) {
flags[score_equal] = true;
return 0;
}
kswr_t best_alignment_info;
if(reverse_return.score > forward_return.score) {
best_alignment_info = reverse_return;
flags[reversed] = true;
} else {
best_alignment_info = forward_return;
}
int min_query_seq_coverage = static_cast<int>(slave_dr_length / 2);
// this is not the way you are supposed to use goto
// actually you're never supposed to use goto
// but you know what, I don't care!
// http://xkcd.com/292/
if(min_query_seq_coverage > best_alignment_info.score) {
logWarn("Alignment Warning: Slave Alignment Failure",4);
logWarn("\tfailed query coverage test", 4);
logWarn("\trequired: "<<min_query_seq_coverage, 4);
goto FAILED;
}
if(best_alignment_info.score < AL_minAlignmentScore) {
logWarn("Alignment Warning: Slave Alignment Failure",4);
logWarn("\tfailed minimum score test", 4);
goto FAILED;
}
//if (best_alignment_info.qb != 0 && best_alignment_info.qe != slave_dr_length - 1) {
// logWarn("Alignment Warning: Slave Alignment Failure",4);
// logWarn("\tfailed internal only test", 4);
// goto FAILED;
//}
//std::cerr << best_alignment_info.tb - best_alignment_info.qb << " : ";
//this->printAlignment(best_alignment_info, slaveDR, std::cerr);
return best_alignment_info.tb - best_alignment_info.qb;
FAILED:
logWarn("\tmaster: "<<mStringCheck->getString(AL_masterDRToken) , 4);
logWarn("\ttb: "<< best_alignment_info.tb, 4);
logWarn("\tte: "<< best_alignment_info.te+1, 4);
logWarn("\tslave: "<< slaveDR, 4);
logWarn("\tqb: "<< best_alignment_info.qb, 4);
logWarn("\tqe: "<< best_alignment_info.qe+1, 4);
logWarn("\tscore: "<< best_alignment_info.score, 4);
logWarn("\t2nd-score: "<< best_alignment_info.score2, 4);
logWarn("\t2nd-te: "<< best_alignment_info.te2, 4);
logWarn("\toffset: "<<best_alignment_info.tb - best_alignment_info.qb,4);
logWarn("******", 4);
flags[failed] = true;
return 0;
}
static aln_t align_read_against_one(kseq_t *target, const int read_len,
uint8_t *read_num, kswq_t **qry,
const align_config_t *conf,
const int min_score) {
uint8_t *ref_num = calloc(target->seq.l, sizeof(uint8_t));
for (size_t k = 0; k < target->seq.l; ++k)
ref_num[k] = conf->table[(int)target->seq.s[k]];
aln_t aln;
aln.cigar = NULL;
aln.loc = ksw_align(read_len, read_num, target->seq.l, ref_num, conf->m,
conf->mat, conf->gap_o, conf->gap_e, KSW_XSTART, qry);
aln.target_name = target->name.s;
if (aln.loc.score < min_score) {
free(ref_num);
return aln;
}
ksw_global(aln.loc.qe - aln.loc.qb + 1, &read_num[aln.loc.qb],
aln.loc.te - aln.loc.tb + 1, &ref_num[aln.loc.tb], conf->m,
conf->mat, conf->gap_o, conf->gap_e, conf->bandwidth, &aln.n_cigar,
&aln.cigar);
aln.nm = 0;
size_t qi = aln.loc.qb, ri = aln.loc.tb;
for (int k = 0; k < aln.n_cigar; k++) {
const int32_t oplen = bam_cigar_oplen(aln.cigar[k]),
optype = bam_cigar_type(aln.cigar[k]);
if (optype & 3) { // consumes both - check for mismatches
for (int j = 0; j < oplen; j++) {
if (UNLIKELY(read_num[qi + j] != ref_num[ri + j]))
aln.nm++;
}
} else {
aln.nm += oplen;
}
if (optype & 1)
qi += oplen;
if (optype & 2)
ri += oplen;
}
free(ref_num);
/* size_t cigar_len = aln.loc.qb; */
/* for (int c = 0; c < aln.n_cigar; c++) { */
/* int32_t length = (0xfffffff0 & *(aln.cigar + c)) >> 4; */
/* cigar_len += length; */
/* } */
/* cigar_len += read_len - aln.loc.qe - 1; */
/* if(cigar_len != (size_t)read_len) { */
/* /\* printf("[ig_align] Error: cigar length (score %d) not equal to read length for XXX (target %s): %zu vs %d\n", aln.loc.score, target->name.s, cigar_len, read_len); *\/ */
/* // NOTE: */
/* // It is *really* *f*****g* *scary* that it's spitting out cigars that are not the same length as the query sequence. */
/* // Nonetheless, fixing it seems to involve delving into the depths of ksw_align() and ksw_global(), which would be very time consuming, and the length discrepancy seems to ony appear in very poor matches. */
/* // I.e., poor enough that we will subsequently ignore them in partis/python/waterer.py, so it seems to not screw anything up downstream to just set the length-discrepant matches' scores to zero, such that ig-sw doesn't write them to its sam output. */
/* // Note also that it is not always the lowest- or highest-scoring matches that have discrepant lengths (i.e. setting their scores to zero promotes matches swith poorer scores, but which do not have discrepant lengths. */
/* /\* aln.loc.score = 0; *\/ */
/* aln.cigar = NULL; */
/* } */
return aln;
}
uint8_t* slave_dr_reverse = new uint8_t[slave_dr_length+1];
prepareSlaveForAlignment(slaveDR, slave_dr_forward, slave_dr_reverse);
// query profile
kswq_t *slave_forward_query_profile = 0;
kswq_t *slave_reverse_query_profile = 0;
// alignment of slave against master
kswr_t forward_return = ksw_align(slave_dr_length,
slave_dr_forward,
AL_masterDRLength,
AL_masterDR,
5,
AL_scoringMatrix,
AL_gapOpening,
AL_gapExtension,
AL_xtra,
&slave_forward_query_profile);
kswr_t reverse_return = ksw_align(slave_dr_length,
slave_dr_reverse,
AL_masterDRLength,
AL_masterDR,
5,
AL_scoringMatrix,
AL_gapOpening,
AL_gapExtension,
AL_xtra,