void bwa_cal_pac_pos(const char *prefix, int n_seqs, bwa_seq_t *seqs, int max_mm, float fnr)
{
int i;
char str[1024];
bwt_t *bwt;
// load forward SA
strcpy(str, prefix); strcat(str, ".bwt"); bwt = bwt_restore_bwt(str);
strcpy(str, prefix); strcat(str, ".sa"); bwt_restore_sa(str, bwt);
for (i = 0; i != n_seqs; ++i) {
bwa_seq_t *p = seqs + i;
int max_diff = fnr > 0.0? bwa_cal_maxdiff(p->len, BWA_AVG_ERR, fnr) : max_mm;
if ((p->type == BWA_TYPE_UNIQUE || p->type == BWA_TYPE_REPEAT) && p->strand) { // reverse strand only
p->pos = bwt_sa(bwt, p->sa);
p->seQ = p->mapQ = bwa_approx_mapQ(p, max_diff);
}
}
bwt_destroy(bwt);
// load reverse BWT and SA
strcpy(str, prefix); strcat(str, ".rbwt"); bwt = bwt_restore_bwt(str);
strcpy(str, prefix); strcat(str, ".rsa"); bwt_restore_sa(str, bwt);
for (i = 0; i != n_seqs; ++i) {
bwa_seq_t *p = seqs + i;
int max_diff = fnr > 0.0? bwa_cal_maxdiff(p->len, BWA_AVG_ERR, fnr) : max_mm;
if ((p->type == BWA_TYPE_UNIQUE || p->type == BWA_TYPE_REPEAT) && !p->strand) { // forward strand only
/* NB: For gapped alignment, p->pos may not be correct,
* which will be fixed in refine_gapped_core(). This
* line also determines the way "x" is calculated in
* refine_gapped_core() when (ext < 0 && is_end == 0). */
p->pos = bwt->seq_len - (bwt_sa(bwt, p->sa) + p->len);
p->seQ = p->mapQ = bwa_approx_mapQ(p, max_diff);
}
}
bwt_destroy(bwt);
}
void bwa_cal_sa_reg_gap(int tid, bwt_t *const bwt[2], int n_seqs, bwa_seq_t *seqs, const gap_opt_t *opt)
{
int i, max_l = 0, max_len;
gap_stack_t *stack;
bwt_width_t *w[2], *seed_w[2];
const ubyte_t *seq[2];
gap_opt_t local_opt = *opt;
// initiate priority stack
for (i = max_len = 0; i != n_seqs; ++i)
if (seqs[i].len > max_len) max_len = seqs[i].len;
if (opt->fnr > 0.0) local_opt.max_diff = bwa_cal_maxdiff(max_len, BWA_AVG_ERR, opt->fnr);
if (local_opt.max_diff < local_opt.max_gapo) local_opt.max_gapo = local_opt.max_diff;
stack = gap_init_stack(local_opt.max_diff, local_opt.max_gapo, local_opt.max_gape, &local_opt);
seed_w[0] = (bwt_width_t*)calloc(opt->seed_len+1, sizeof(bwt_width_t));
seed_w[1] = (bwt_width_t*)calloc(opt->seed_len+1, sizeof(bwt_width_t));
w[0] = w[1] = 0;
for (i = 0; i != n_seqs; ++i) {
bwa_seq_t *p = seqs + i;
#ifdef HAVE_PTHREAD
if (opt->n_threads > 1) {
pthread_mutex_lock(&g_seq_lock);
if (p->tid < 0) { // unassigned
int j;
for (j = i; j < n_seqs && j < i + THREAD_BLOCK_SIZE; ++j)
seqs[j].tid = tid;
} else if (p->tid != tid) {
pthread_mutex_unlock(&g_seq_lock);
continue;
}
pthread_mutex_unlock(&g_seq_lock);
}
#endif
p->sa = 0; p->type = BWA_TYPE_NO_MATCH; p->c1 = p->c2 = 0; p->n_aln = 0; p->aln = 0;
seq[0] = p->seq; seq[1] = p->rseq;
if (max_l < p->len) {
max_l = p->len;
w[0] = (bwt_width_t*)realloc(w[0], (max_l + 1) * sizeof(bwt_width_t));
w[1] = (bwt_width_t*)realloc(w[1], (max_l + 1) * sizeof(bwt_width_t));
memset(w[0], 0, (max_l + 1) * sizeof(bwt_width_t));
memset(w[1], 0, (max_l + 1) * sizeof(bwt_width_t));
}
bwt_cal_width(bwt[0], p->len, seq[0], w[0]);
bwt_cal_width(bwt[1], p->len, seq[1], w[1]);
if (opt->fnr > 0.0) local_opt.max_diff = bwa_cal_maxdiff(p->len, BWA_AVG_ERR, opt->fnr);
local_opt.seed_len = opt->seed_len < p->len? opt->seed_len : 0x7fffffff;
if (p->len > opt->seed_len) {
bwt_cal_width(bwt[0], opt->seed_len, seq[0] + (p->len - opt->seed_len), seed_w[0]);
bwt_cal_width(bwt[1], opt->seed_len, seq[1] + (p->len - opt->seed_len), seed_w[1]);
}
// core function
p->aln = bwt_match_gap(bwt, p->len, seq, w, p->len <= opt->seed_len? 0 : seed_w, &local_opt, &p->n_aln, stack);
// store the alignment
free(p->name); free(p->seq); free(p->rseq); free(p->qual);
p->name = 0; p->seq = p->rseq = p->qual = 0;
}
free(seed_w[0]); free(seed_w[1]);
free(w[0]); free(w[1]);
gap_destroy_stack(stack);
}
void mybwa_cal_sa_reg_gap(int tid, bwt_t *const bwt[2], int n_seqs, bwa_seq_t *seqs, const gap_opt_t *opt)
{
int i, max_l = 0, max_len;
gap_stack_t *stack;
bwt_width_t *w[2], *seed_w[2];
const ubyte_t *seq[2];
gap_opt_t local_opt = *opt;
// no mismatches or gaps in seed
local_opt.max_gape=0 ;
local_opt.max_gapo=0 ;
local_opt.max_seed_diff=0 ;
local_opt.max_diff=0 ;
// initiate priority stack
for (i = max_len = 0; i != n_seqs; ++i)
if (seqs[i].len > max_len) max_len = seqs[i].len;
if (opt->fnr > 0.0) local_opt.max_diff = bwa_cal_maxdiff(max_len, BWA_AVG_ERR, opt->fnr);
if (local_opt.max_diff < local_opt.max_gapo) local_opt.max_gapo = local_opt.max_diff;
stack = gap_init_stack(local_opt.max_diff, local_opt.max_gapo, local_opt.max_gape, &local_opt);
seed_w[0] = (bwt_width_t*)calloc(opt->seed_len+1, sizeof(bwt_width_t));
seed_w[1] = (bwt_width_t*)calloc(opt->seed_len+1, sizeof(bwt_width_t));
w[0] = w[1] = 0;
for (i = 0; i != n_seqs; ++i) {
bwa_seq_t *p = seqs + i;
p->sa = 0; p->type = BWA_TYPE_NO_MATCH; p->c1 = p->c2 = 0; p->n_aln = 0; p->aln = 0;
seq[0] = p->seq; seq[1] = p->rseq;
if (max_l < p->len) {
max_l = p->len;
w[0] = (bwt_width_t*)realloc(w[0], (max_l + 1) * sizeof(bwt_width_t));
w[1] = (bwt_width_t*)realloc(w[1], (max_l + 1) * sizeof(bwt_width_t));
memset(w[0], 0, (max_l + 1) * sizeof(bwt_width_t));
memset(w[1], 0, (max_l + 1) * sizeof(bwt_width_t));
}
bwt_cal_width(bwt[0], p->len, seq[0], w[0]);
bwt_cal_width(bwt[1], p->len, seq[1], w[1]);
//fprintf(stdout, "w[0]=%i, w[1]=%i\n", w[0]->w, w[1]->w) ;
if (opt->fnr > 0.0) local_opt.max_diff = bwa_cal_maxdiff(p->len, BWA_AVG_ERR, opt->fnr);
local_opt.seed_len = opt->seed_len < p->len? opt->seed_len : 0x7fffffff;
if (p->len > opt->seed_len) {
bwt_cal_width(bwt[0], opt->seed_len, seq[0] + (p->len - opt->seed_len), seed_w[0]);
bwt_cal_width(bwt[1], opt->seed_len, seq[1] + (p->len - opt->seed_len), seed_w[1]);
}
// core function
p->aln = bwt_match_gap(bwt, p->len, seq, w, p->len <= opt->seed_len? 0 : seed_w, &local_opt, &p->n_aln, stack);
// store the alignment
free(p->name); free(p->seq); free(p->rseq); free(p->qual);
p->name = 0; p->seq = p->rseq = p->qual = 0;
}
free(seed_w[0]); free(seed_w[1]);
free(w[0]); free(w[1]);
gap_destroy_stack(stack);
}
void bwa_cal_pac_pos_core(const bwt_t *bwt, bwa_seq_t *seq, const int max_mm, const float fnr, const int *g_log_n)
{
int max_diff;
if (seq->type != BWA_TYPE_UNIQUE && seq->type != BWA_TYPE_REPEAT) return;
max_diff = fnr > 0.0? bwa_cal_maxdiff(seq->len, BWA_AVG_ERR, fnr) : max_mm;
seq->pos = bwt_sa(bwt, seq->sa);
seq->seQ = seq->mapQ = bwa_approx_mapQ(seq, max_diff, g_log_n);
}
void bwa_cal_sa_reg_gap(int tid, bwt_t *const bwt, int n_seqs, bwa_seq_t *seqs, const gap_opt_t *opt)
{
int i, j, max_l = 0, max_len;
gap_stack_t *stack;
bwt_width_t *w, *seed_w;
gap_opt_t local_opt = *opt;
// initiate priority stack
for (i = max_len = 0; i != n_seqs; ++i)
if (seqs[i].len > max_len) max_len = seqs[i].len;
if (opt->fnr > 0.0) local_opt.max_diff = bwa_cal_maxdiff(max_len, BWA_AVG_ERR, opt->fnr);
if (local_opt.max_diff < local_opt.max_gapo) local_opt.max_gapo = local_opt.max_diff;
stack = gap_init_stack(local_opt.max_diff, local_opt.max_gapo, local_opt.max_gape, &local_opt);
seed_w = (bwt_width_t*)calloc(opt->seed_len+1, sizeof(bwt_width_t));
w = 0;
for (i = 0; i != n_seqs; ++i) {
bwa_seq_t *p = seqs + i;
#ifdef HAVE_PTHREAD
if (i % opt->n_threads != tid) continue;
#endif
p->sa = 0; p->type = BWA_TYPE_NO_MATCH; p->c1 = p->c2 = 0; p->n_aln = 0; p->aln = 0;
if (max_l < p->len) {
max_l = p->len;
w = (bwt_width_t*)realloc(w, (max_l + 1) * sizeof(bwt_width_t));
memset(w, 0, (max_l + 1) * sizeof(bwt_width_t));
}
bwt_cal_width(bwt, p->len, p->seq, w);
if (opt->fnr > 0.0) local_opt.max_diff = bwa_cal_maxdiff(p->len, BWA_AVG_ERR, opt->fnr);
local_opt.seed_len = opt->seed_len < p->len? opt->seed_len : 0x7fffffff;
if (p->len > opt->seed_len)
bwt_cal_width(bwt, opt->seed_len, p->seq + (p->len - opt->seed_len), seed_w);
// core function
for (j = 0; j < p->len; ++j) // we need to complement
p->seq[j] = p->seq[j] > 3? 4 : 3 - p->seq[j];
p->aln = bwt_match_gap(bwt, p->len, p->seq, w, p->len <= opt->seed_len? 0 : seed_w, &local_opt, &p->n_aln, stack);
//fprintf(stderr, "mm=%lld,ins=%lld,del=%lld,gapo=%lld\n", p->aln->n_mm, p->aln->n_ins, p->aln->n_del, p->aln->n_gapo);
// clean up the unused data in the record
free(p->name); free(p->seq); free(p->rseq); free(p->qual);
p->name = 0; p->seq = p->rseq = p->qual = 0;
}
free(seed_w); free(w);
gap_destroy_stack(stack);
}
/**
* Derive the actual position in the read from the given suffix array
* coordinates. Note that the position will be approximate based on
* whether indels appear in the read and whether calculations are
* performed from the start or end of the read.
*/
void bwa_cal_pac_pos_core(const bntseq_t *bns, const bwt_t *bwt, bwa_seq_t *seq, const int max_mm, const float fnr)
{
int max_diff, strand;
if (seq->type != BWA_TYPE_UNIQUE && seq->type != BWA_TYPE_REPEAT) return;
max_diff = fnr > 0.0? bwa_cal_maxdiff(seq->len, BWA_AVG_ERR, fnr) : max_mm;
seq->seQ = seq->mapQ = bwa_approx_mapQ(seq, max_diff);
seq->pos = bwa_sa2pos(bns, bwt, seq->sa, seq->len, &strand);
seq->strand = strand;
seq->seQ = seq->mapQ = bwa_approx_mapQ(seq, max_diff);
}
void bwa_cal_pac_pos_core(const bwt_t *forward_bwt, const bwt_t *reverse_bwt, bwa_seq_t *seq, const int max_mm, const float fnr)
{
int max_diff;
if (seq->type != BWA_TYPE_UNIQUE && seq->type != BWA_TYPE_REPEAT) return;
max_diff = fnr > 0.0? bwa_cal_maxdiff(seq->len, BWA_AVG_ERR, fnr) : max_mm;
if (seq->strand) { // reverse strand only
seq->pos = bwt_sa(forward_bwt, seq->sa);
} else { // forward strand only
/* NB: For gapped alignment, p->pos may not be correct, which
* will be fixed in refine_gapped_core(). This line also
* determines the way "x" is calculated in
* refine_gapped_core() when (ext < 0 && is_end == 0). */
seq->pos = reverse_bwt->seq_len - (bwt_sa(reverse_bwt, seq->sa) + seq->len);
}
seq->seQ = seq->mapQ = bwa_approx_mapQ(seq, max_diff);
}
int bwa_aln(int argc, char *argv[])
{
int c, opte = -1;
gap_opt_t *opt;
opt = gap_init_opt();
while ((c = getopt(argc, argv, "n:o:e:i:d:l:k:cLR:m:t:NM:O:E:q:f:b012IYB:")) >= 0) {
switch (c) {
case 'n':
if (strstr(optarg, ".")) opt->fnr = atof(optarg), opt->max_diff = -1;
else opt->max_diff = atoi(optarg), opt->fnr = -1.0;
break;
case 'o': opt->max_gapo = atoi(optarg); break;
case 'e': opte = atoi(optarg); break;
case 'M': opt->s_mm = atoi(optarg); break;
case 'O': opt->s_gapo = atoi(optarg); break;
case 'E': opt->s_gape = atoi(optarg); break;
case 'd': opt->max_del_occ = atoi(optarg); break;
case 'i': opt->indel_end_skip = atoi(optarg); break;
case 'l': opt->seed_len = atoi(optarg); break;
case 'k': opt->max_seed_diff = atoi(optarg); break;
case 'm': opt->max_entries = atoi(optarg); break;
case 't': opt->n_threads = atoi(optarg); break;
case 'L': opt->mode |= BWA_MODE_LOGGAP; break;
case 'R': opt->max_top2 = atoi(optarg); break;
case 'q': opt->trim_qual = atoi(optarg); break;
case 'c': opt->mode &= ~BWA_MODE_COMPREAD; break;
case 'N': opt->mode |= BWA_MODE_NONSTOP; opt->max_top2 = 0x7fffffff; break;
case 'f': xreopen(optarg, "wb", stdout); break;
case 'b': opt->mode |= BWA_MODE_BAM; break;
case '0': opt->mode |= BWA_MODE_BAM_SE; break;
case '1': opt->mode |= BWA_MODE_BAM_READ1; break;
case '2': opt->mode |= BWA_MODE_BAM_READ2; break;
case 'I': opt->mode |= BWA_MODE_IL13; break;
case 'Y': opt->mode |= BWA_MODE_CFY; break;
case 'B': opt->mode |= atoi(optarg) << 24; break;
default: return 1;
}
}
if (opte > 0) {
opt->max_gape = opte;
opt->mode &= ~BWA_MODE_GAPE;
}
if (optind + 2 > argc) {
fprintf(stderr, "\n");
fprintf(stderr, "Usage: bwa aln [options] <prefix> <in.fq>\n\n");
fprintf(stderr, "Options: -n NUM max #diff (int) or missing prob under %.2f err rate (float) [%.2f]\n",
BWA_AVG_ERR, opt->fnr);
fprintf(stderr, " -o INT maximum number or fraction of gap opens [%d]\n", opt->max_gapo);
fprintf(stderr, " -e INT maximum number of gap extensions, -1 for disabling long gaps [-1]\n");
fprintf(stderr, " -i INT do not put an indel within INT bp towards the ends [%d]\n", opt->indel_end_skip);
fprintf(stderr, " -d INT maximum occurrences for extending a long deletion [%d]\n", opt->max_del_occ);
fprintf(stderr, " -l INT seed length [%d]\n", opt->seed_len);
fprintf(stderr, " -k INT maximum differences in the seed [%d]\n", opt->max_seed_diff);
fprintf(stderr, " -m INT maximum entries in the queue [%d]\n", opt->max_entries);
fprintf(stderr, " -t INT number of threads [%d]\n", opt->n_threads);
fprintf(stderr, " -M INT mismatch penalty [%d]\n", opt->s_mm);
fprintf(stderr, " -O INT gap open penalty [%d]\n", opt->s_gapo);
fprintf(stderr, " -E INT gap extension penalty [%d]\n", opt->s_gape);
fprintf(stderr, " -R INT stop searching when there are >INT equally best hits [%d]\n", opt->max_top2);
fprintf(stderr, " -q INT quality threshold for read trimming down to %dbp [%d]\n", BWA_MIN_RDLEN, opt->trim_qual);
fprintf(stderr, " -f FILE file to write output to instead of stdout\n");
fprintf(stderr, " -B INT length of barcode\n");
// fprintf(stderr, " -c input sequences are in the color space\n");
fprintf(stderr, " -L log-scaled gap penalty for long deletions\n");
fprintf(stderr, " -N non-iterative mode: search for all n-difference hits (slooow)\n");
fprintf(stderr, " -I the input is in the Illumina 1.3+ FASTQ-like format\n");
fprintf(stderr, " -b the input read file is in the BAM format\n");
fprintf(stderr, " -0 use single-end reads only (effective with -b)\n");
fprintf(stderr, " -1 use the 1st read in a pair (effective with -b)\n");
fprintf(stderr, " -2 use the 2nd read in a pair (effective with -b)\n");
fprintf(stderr, " -Y filter Casava-filtered sequences\n");
fprintf(stderr, "\n");
return 1;
}
if (opt->fnr > 0.0) {
int i, k;
for (i = 17, k = 0; i <= 250; ++i) {
int l = bwa_cal_maxdiff(i, BWA_AVG_ERR, opt->fnr);
if (l != k) fprintf(stderr, "[bwa_aln] %dbp reads: max_diff = %d\n", i, l);
k = l;
}
}
bwa_aln_core(argv[optind], argv[optind+1], opt);
free(opt);
return 0;
}