bam_fillmd1_core(b, ref, is_equal, 0);
}
-int bam_cap_mapQ(bam1_t *b, char *ref)
+int bam_cap_mapQ(bam1_t *b, char *ref, int thres)
{
uint8_t *seq = bam1_seq(b), *qual = bam1_qual(b);
uint32_t *cigar = bam1_cigar(b);
bam1_core_t *c = &b->core;
int i, x, y, mm, q, len, clip_l, clip_q;
double t;
-
+ if (thres < 0) thres = 40; // set the default
mm = q = len = clip_l = clip_q = 0;
for (i = y = 0, x = c->pos; i < c->n_cigar; ++i) {
int j, l = cigar[i]>>4, op = cigar[i]&0xf;
for (i = 0, t = 1; i < mm; ++i)
t *= (double)len / (i+1);
t = q - 4.343 * log(t) + clip_q / 5.;
- if (t > 40) return -1;
+ if (t > thres) return -1;
if (t < 0) t = 0;
- t = sqrt((40 - t) / 40) * 40;
- fprintf(stderr, "%s %lf %d\n", bam1_qname(b), t, q);
+ t = sqrt((thres - t) / thres) * thres;
+// fprintf(stderr, "%s %lf %d\n", bam1_qname(b), t, q);
return (int)(t + .499);
}
-// local realignment
-
-#define MIN_REF_LEN 10
-#define MIN_BAND_WIDTH 11
-
-int bam_realn(bam1_t *b, const char *ref)
+int bam_prob_realn(bam1_t *b, const char *ref)
{
- int k, score, q[2], r[2], kk[2], kl[2], x, y, max, j, n_cigar, endx = -1;
+ int k, i, bw, x, y, yb, ye, xb, xe;
uint32_t *cigar = bam1_cigar(b);
- uint8_t *seq = bam1_seq(b);
bam1_core_t *c = &b->core;
- q[0] = q[1] = r[0] = r[1] = kk[0] = kk[1] = kl[0] = kl[1] = -1;
- // find the right boundary
- for (k = 0, score = max = 0, x = c->pos, y = 0; k < c->n_cigar; ++k) {
- int op = cigar[k]&0xf;
- int ol = cigar[k]>>4;
+ ka_probpar_t conf = ka_probpar_def;
+ // find the start and end of the alignment
+ if (c->flag & BAM_FUNMAP) return -1;
+ x = c->pos, y = 0, yb = ye = xb = xe = -1;
+ for (k = 0; k < c->n_cigar; ++k) {
+ int op, l;
+ op = cigar[k]&0xf; l = cigar[k]>>4;
if (op == BAM_CMATCH) {
- for (j = 0; j < ol; ++j) {
- int c1, c2, z = y + j;
- c1 = bam_nt16_nt4_table[bam1_seqi(seq, z)];
- if (ref[x+j] == 0) return -1;
- c2 = bam_nt16_nt4_table[(int)bam_nt16_table[(int)ref[x+j]]];
- if (c1 < 3 && c2 < 3)
- score += c1 == c2? 5 : -4;
- if (score < 0) score = 0;
- if (score > max) max = score, q[1] = z, r[1] = x+j, kk[1] = k, kl[1] = j + 1;
- }
- x += ol; y += ol;
- } else if (op == BAM_CINS) {
- score -= 4 - ol * 3;
- y += ol;
- if (score < 0) score = 0;
- } else if (op == BAM_CDEL) {
- score -= 4 - ol * 3;
- x += ol;
- if (score < 0) score = 0;
- } else if (op == BAM_CSOFT_CLIP) y += ol;
- else if (op == BAM_CREF_SKIP) x += ol;
- }
- if (score < 0) return -1; // no high scoring segments
- endx = x - 1;
- // find the left boundary
- for (k = c->n_cigar - 1, score = max = 0, x = x-1, y = y-1; k >= 0; --k) {
- int op = cigar[k]&0xf;
- int ol = cigar[k]>>4;
- if (op == BAM_CMATCH) {
- for (j = 0; j < ol; ++j) {
- int c1, c2, z = y - j;
- c1 = bam_nt16_nt4_table[bam1_seqi(seq, z)];
- if (ref[x+j] == 0) return -1;
- c2 = bam_nt16_nt4_table[(int)bam_nt16_table[(int)ref[x-j]]];
- if (c1 < 3 && c2 < 3)
- score += c1 == c2? 5 : -4;
- if (score < 0) score = 0;
- if (score > max) max = score, q[0] = z, r[0] = x-j, kk[0] = k, kl[0] = j + 1;
- }
- x -= ol; y -= ol;
- } else if (op == BAM_CINS) {
- score -= 4 - ol * 3;
- y -= ol;
- if (score < 0) score = 0;
- } else if (op == BAM_CDEL) {
- score -= 4 - ol * 3;
- x -= ol;
- if (score < 0) score = 0;
- } else if (op == BAM_CSOFT_CLIP) y -= ol;
- else if (op == BAM_CREF_SKIP) x -= ol;
- }
- if (q[1] - q[0] < 15) return -1; // the high-scoring segment is too short
- // modify CIGAR
- n_cigar = 0;
- cigar = calloc(c->n_cigar + 4, 4);
- if (q[0] != 0) cigar[n_cigar++] = (uint32_t)q[0]<<4 | BAM_CSOFT_CLIP;
- if (r[0] != c->pos) cigar[n_cigar++] = (uint32_t)(r[0] - c->pos)<<4 | BAM_CREF_SKIP;
- if (kk[0] == kk[1]) {
- cigar[n_cigar++] = (uint32_t)(kl[0] + kl[1] - (bam1_cigar(b)[kk[0]]>>4))<<4 | BAM_CMATCH;
- } else {
- cigar[n_cigar++] = (uint32_t)kl[0]<<4 | BAM_CMATCH;
- for (k = kk[0] + 1; k < kk[1]; ++k)
- cigar[n_cigar++] = bam1_cigar(b)[k];
- cigar[n_cigar++] = (uint32_t)kl[1]<<4 | BAM_CMATCH; // FIXME: add ref_skip after this line
- }
- if (q[1] + 1 != c->l_qseq)
- cigar[n_cigar++] = (uint32_t)(c->l_qseq - q[1] - 1)<<4 | BAM_CSOFT_CLIP;
- // copy over the alignment
- if (4 * (n_cigar - (int)c->n_cigar) + b->data_len > b->m_data) { // enlarge b->data
- b->m_data = 4 * (n_cigar - (int)c->n_cigar) + b->data_len;
- kroundup32(b->m_data);
- b->data = realloc(b->data, b->m_data);
+ if (yb < 0) yb = y;
+ if (xb < 0) xb = x;
+ ye = y + l; xe = x + l;
+ x += l; y += l;
+ } else if (op == BAM_CSOFT_CLIP || op == BAM_CINS) y += l;
+ else if (op == BAM_CDEL) x += l;
+ else if (op == BAM_CREF_SKIP) return -1;
}
- if (n_cigar != (int)c->n_cigar) { // move data
- memmove(b->data + c->l_qname + 4 * n_cigar, bam1_seq(b), b->data_len - c->l_qname - 4 * c->n_cigar);
- b->data_len += 4 * (n_cigar - (int)c->n_cigar);
+ // set bandwidth and the start and the end
+ bw = 7;
+ if (abs((xe - xb) - (ye - yb)) > bw)
+ bw = abs((xe - xb) - (ye - yb)) + 3;
+ conf.bw = bw;
+ xb -= yb + bw/2; if (xb < 0) xb = 0;
+ xe += c->l_qseq - ye + bw/2;
+ if (xe - xb - c->l_qseq > bw)
+ xb += (xe - xb - c->l_qseq - bw) / 2, xe -= (xe - xb - c->l_qseq - bw) / 2;
+ { // glocal
+ uint8_t *s, *r, *q, *seq = bam1_seq(b), *qual = bam1_qual(b);
+ int *state;
+ s = calloc(c->l_qseq, 1);
+ for (i = 0; i < c->l_qseq; ++i) s[i] = bam_nt16_nt4_table[bam1_seqi(seq, i)];
+ r = calloc(xe - xb, 1);
+ for (i = xb; i < xe; ++i)
+ r[i-xb] = bam_nt16_nt4_table[bam_nt16_table[(int)ref[i]]];
+ state = calloc(c->l_qseq, sizeof(int));
+ q = calloc(c->l_qseq, 1);
+ ka_prob_glocal(r, xe-xb, s, c->l_qseq, qual, &conf, state, q);
+ for (k = 0, x = c->pos, y = 0; k < c->n_cigar; ++k) {
+ int op = cigar[k]&0xf, l = cigar[k]>>4;
+ if (op == BAM_CMATCH) {
+ for (i = y; i < y + l; ++i) {
+ if ((state[i]&3) != 0 || state[i]>>2 != x - xb + (i - y)) qual[i] = 0;
+ else qual[i] = qual[i] < q[i]? qual[i] : q[i];
+ }
+ x += l; y += l;
+ } else if (op == BAM_CSOFT_CLIP || op == BAM_CINS) y += l;
+ else if (op == BAM_CDEL) x += l;
+ }
+ free(s); free(r); free(q); free(state);
}
- memcpy(bam1_cigar(b), cigar, n_cigar * 4);
- c->n_cigar = n_cigar;
- free(cigar);
return 0;
}
int bam_fillmd(int argc, char *argv[])
{
- int c, is_equal = 0, tid = -2, ret, len, is_bam_out, is_sam_in, is_uncompressed, max_nm = 0, is_realn, is_capQ;
+ int c, is_equal = 0, tid = -2, ret, len, is_bam_out, is_sam_in, is_uncompressed, max_nm = 0, is_realn, capQ = 0;
samfile_t *fp, *fpout = 0;
faidx_t *fai;
char *ref = 0, mode_w[8], mode_r[8];
bam1_t *b;
- is_bam_out = is_sam_in = is_uncompressed = is_realn = is_capQ = 0;
+ is_bam_out = is_sam_in = is_uncompressed = is_realn = 0;
mode_w[0] = mode_r[0] = 0;
strcpy(mode_r, "r"); strcpy(mode_w, "w");
- while ((c = getopt(argc, argv, "reubSCn:")) >= 0) {
+ while ((c = getopt(argc, argv, "reubSC:n:")) >= 0) {
switch (c) {
case 'r': is_realn = 1; break;
case 'e': is_equal = 1; break;
case 'u': is_uncompressed = is_bam_out = 1; break;
case 'S': is_sam_in = 1; break;
case 'n': max_nm = atoi(optarg); break;
- case 'C': is_capQ = 1; break;
+ case 'C': capQ = atoi(optarg); break;
default: fprintf(stderr, "[bam_fillmd] unrecognized option '-%c'\n", c); return 1;
}
}
fprintf(stderr, "[bam_fillmd] fail to find sequence '%s' in the reference.\n",
fp->header->target_name[tid]);
}
- if (is_realn) bam_realn(b, ref);
- if (is_capQ) {
- int q = bam_cap_mapQ(b, ref);
+// if (is_realn) bam_realn(b, ref);
+ if (is_realn) bam_prob_realn(b, ref);
+ if (capQ > 10) {
+ int q = bam_cap_mapQ(b, ref, capQ);
if (b->core.qual > q) b->core.qual = q;
}
if (ref) bam_fillmd1_core(b, ref, is_equal, max_nm);