7 #include "bam_maqcns.h"
12 typedef int *indel_list_t;
13 KHASH_MAP_INIT_INT64(64, indel_list_t)
15 #define BAM_PLF_SIMPLE 0x01
16 #define BAM_PLF_CNS 0x02
17 #define BAM_PLF_INDEL_ONLY 0x04
18 #define BAM_PLF_GLF 0x08
19 #define BAM_PLF_VAR_ONLY 0x10
20 #define BAM_PLF_2ND 0x20
21 #define BAM_PLF_RANBASE 0x40
22 #define BAM_PLF_1STBASE 0x80
23 #define BAM_PLF_ALLBASE 0x100
24 #define BAM_PLF_READPOS 0x200
29 bam_maqindel_opt_t *ido;
33 int tid, len, last_pos;
35 int max_depth; // for indel calling, ignore reads with the depth too high. 0 for unlimited
37 glfFile fp_glf; // for glf output only
40 char **__bam_get_lines(const char *fn, int *_n);
41 void bam_init_header_hash(bam_header_t *header);
42 int32_t bam_get_tid(const bam_header_t *header, const char *seq_name);
44 static khash_t(64) *load_pos(const char *fn, bam_header_t *h)
47 int i, j, n, *fields, max_fields;
49 bam_init_header_hash(h);
50 list = __bam_get_lines(fn, &n);
52 max_fields = 0; fields = 0;
53 for (i = 0; i < n; ++i) {
55 int chr, n_fields, ret;
58 n_fields = ksplit_core(str, 0, &max_fields, &fields);
59 if (n_fields < 2) continue;
60 chr = bam_get_tid(h, str + fields[0]);
62 fprintf(stderr, "[load_pos] unknown reference sequence name: %s\n", str + fields[0]);
65 x = (uint64_t)chr << 32 | (atoi(str + fields[1]) - 1);
66 k = kh_put(64, hash, x, &ret);
68 fprintf(stderr, "[load_pos] position %s:%s has been loaded.\n", str+fields[0], str+fields[1]);
74 for (j = 2; j < n_fields; ++j) {
75 char *s = str + fields[j];
76 if ((*s != '+' && *s != '-') || !isdigit(s[1])) break;
78 if (j > 2) { // update kh_val()
80 q = kh_val(hash, k) = (int*)calloc(j - 1, sizeof(int));
81 q[0] = j - 2; z = j; y = 1;
82 for (j = 2; j < z; ++j)
83 q[y++] = atoi(str + fields[j]);
88 free(list); free(fields);
92 // an analogy to pileup_func() below
93 static int glt3_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *pu, void *data)
95 pu_data_t *d = (pu_data_t*)data;
96 bam_maqindel_ret_t *r = 0;
97 int rb, *proposed_indels = 0;
102 fprintf(stderr, "[glt3_func] reference sequence is required for generating GLT. Abort!\n");
105 if (d->hash) { // only output a list of sites
106 khint_t k = kh_get(64, d->hash, (uint64_t)tid<<32|pos);
107 if (k == kh_end(d->hash)) return 0;
108 proposed_indels = kh_val(d->hash, k);
111 if (d->fai && (int)tid != d->tid) {
112 if (d->ref) { // then write the end mark
113 g3->rtype = GLF3_RTYPE_END;
114 glf3_write1(d->fp_glf, g3);
116 glf3_ref_write(d->fp_glf, d->h->target_name[tid], d->h->target_len[tid]); // write reference
118 d->ref = fai_fetch(d->fai, d->h->target_name[tid], &d->len);
122 rb = (d->ref && (int)pos < d->len)? d->ref[pos] : 'N';
123 g = bam_maqcns_glfgen(n, pu, bam_nt16_table[rb], d->c);
124 memcpy(g3, g, sizeof(glf1_t));
125 g3->rtype = GLF3_RTYPE_SUB;
126 g3->offset = pos - d->last_pos;
128 glf3_write1(d->fp_glf, g3);
130 int m = (!d->max_depth || d->max_depth>n) ? n : d->max_depth;
132 r = bam_maqindel(m, pos, d->ido, pu, d->ref, proposed_indels[0], proposed_indels+1);
133 else r = bam_maqindel(m, pos, d->ido, pu, d->ref, 0, 0);
135 if (r) { // then write indel line
136 int het = 3 * n, min;
138 if (min > r->gl[0]) min = r->gl[0];
139 if (min > r->gl[1]) min = r->gl[1];
141 g3->rtype = GLF3_RTYPE_INDEL;
142 memset(g3->lk, 0, 10);
143 g3->lk[0] = r->gl[0] - min < 255? r->gl[0] - min : 255;
144 g3->lk[1] = r->gl[1] - min < 255? r->gl[1] - min : 255;
145 g3->lk[2] = het - min < 255? het - min : 255;
147 g3->indel_len[0] = r->indel1;
148 g3->indel_len[1] = r->indel2;
149 g3->min_lk = min < 255? min : 255;
150 g3->max_len = (abs(r->indel1) > abs(r->indel2)? abs(r->indel1) : abs(r->indel2)) + 1;
151 g3->indel_seq[0] = strdup(r->s[0]+1);
152 g3->indel_seq[1] = strdup(r->s[1]+1);
153 glf3_write1(d->fp_glf, g3);
154 bam_maqindel_ret_destroy(r);
161 static void pileup_seq(const bam_pileup1_t *p, int pos, int ref_len, const char *ref)
163 if (p->is_head) printf("^%c", p->b->core.qual > 93? 126 : p->b->core.qual + 33);
165 int j, rb, c = bam_nt16_rev_table[bam1_seqi(bam1_seq(p->b), p->qpos)];
166 rb = (ref && pos < ref_len)? ref[pos] : 'N';
167 if (c == '=' || toupper(c) == toupper(rb)) c = bam1_strand(p->b)? ',' : '.';
168 else c = bam1_strand(p->b)? tolower(c) : toupper(c);
171 printf("+%d", p->indel);
172 for (j = 1; j <= p->indel; ++j) {
173 c = bam_nt16_rev_table[bam1_seqi(bam1_seq(p->b), p->qpos + j)];
174 putchar(bam1_strand(p->b)? tolower(c) : toupper(c));
176 } else if (p->indel < 0) {
177 printf("%d", p->indel);
178 for (j = 1; j <= -p->indel; ++j) {
179 c = (ref && (int)pos+j < ref_len)? ref[pos+j] : 'N';
180 putchar(bam1_strand(p->b)? tolower(c) : toupper(c));
184 if (p->is_tail) putchar('$');
187 static int pileup_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *pu, void *data)
189 pu_data_t *d = (pu_data_t*)data;
190 bam_maqindel_ret_t *r = 0;
191 int i, rb, rms_mapq = -1, *proposed_indels = 0;
195 // if GLF is required, suppress -c completely
196 if (d->format & BAM_PLF_GLF) return glt3_func(tid, pos, n, pu, data);
197 // if d->hash is initialized, only output the sites in the hash table
199 khint_t k = kh_get(64, d->hash, (uint64_t)tid<<32|pos);
200 if (k == kh_end(d->hash)) return 0;
201 proposed_indels = kh_val(d->hash, k);
203 // update d->ref if necessary
204 if (d->fai && (int)tid != d->tid) {
206 d->ref = faidx_fetch_seq(d->fai, d->h->target_name[tid], 0, 0x7fffffff, &d->len);
209 rb = (d->ref && (int)pos < d->len)? d->ref[pos] : 'N';
210 // when the indel-only mode is asked for, return if no reads mapped with indels
211 if (d->format & BAM_PLF_INDEL_ONLY) {
212 for (i = 0; i < n; ++i)
213 if (pu[i].indel != 0) break;
214 if (i == n) return 0;
216 // call the consensus and indel
217 if (d->format & BAM_PLF_CNS) { // call consensus
218 if (d->format & (BAM_PLF_RANBASE|BAM_PLF_1STBASE)) { // use a random base or the 1st base as the consensus call
219 const bam_pileup1_t *p = (d->format & BAM_PLF_1STBASE)? pu : pu + (int)(drand48() * n);
220 int q = bam1_qual(p->b)[p->qpos];
221 int mapQ = p->b->core.qual < d->c->cap_mapQ? p->b->core.qual : d->c->cap_mapQ;
222 uint32_t b = bam1_seqi(bam1_seq(p->b), p->qpos);
223 cns = b<<28 | 0xf<<24 | mapQ<<16 | q<<8;
224 } else if (d->format & BAM_PLF_ALLBASE) { // collapse all bases
227 for (i = 0; i < n; ++i) {
228 const bam_pileup1_t *p = pu + i;
229 int q = p->b->core.qual < d->c->cap_mapQ? p->b->core.qual : d->c->cap_mapQ;
230 b |= bam1_seqi(bam1_seq(p->b), p->qpos);
233 rmsQ = (uint64_t)(sqrt((double)rmsQ / n) + .499);
234 cns = b<<28 | 0xf<<24 | rmsQ<<16 | 60<<8;
235 } else cns = bam_maqcns_call(n, pu, d->c);
237 if ((d->format & (BAM_PLF_CNS|BAM_PLF_INDEL_ONLY)) && d->ref && pos < d->len) { // call indels
238 int m = (!d->max_depth || d->max_depth>n) ? n : d->max_depth;
239 if (proposed_indels) // the first element gives the size of the array
240 r = bam_maqindel(m, pos, d->ido, pu, d->ref, proposed_indels[0], proposed_indels+1);
241 else r = bam_maqindel(m, pos, d->ido, pu, d->ref, 0, 0);
243 // when only variant sites are asked for, test if the site is a variant
244 if ((d->format & BAM_PLF_CNS) && (d->format & BAM_PLF_VAR_ONLY)) {
245 if (!(bam_nt16_table[rb] != 15 && cns>>28 != bam_nt16_table[rb])) { // not a SNP
246 if (!(r && (r->gt == 2 || strcmp(r->s[r->gt], "*")))) { // not an indel
247 if (r) bam_maqindel_ret_destroy(r);
252 // print the first 3 columns
253 printf("%s\t%d\t%c\t", d->h->target_name[tid], pos + 1, rb);
254 // print consensus information if required
255 if (d->format & BAM_PLF_CNS) {
256 int ref_q, rb4 = bam_nt16_table[rb];
258 if (rb4 != 15 && cns>>28 != 15 && cns>>28 != rb4) { // a SNP
259 ref_q = ((cns>>24&0xf) == rb4)? cns>>8&0xff : (cns>>8&0xff) + (cns&0xff);
260 if (ref_q > 255) ref_q = 255;
262 rms_mapq = cns>>16&0xff;
263 printf("%c\t%d\t%d\t%d\t", bam_nt16_rev_table[cns>>28], cns>>8&0xff, ref_q, rms_mapq);
265 // print pileup sequences
267 rms_aux = 0; // we need to recalculate rms_mapq when -c is not flagged on the command line
268 for (i = 0; i < n; ++i) {
269 const bam_pileup1_t *p = pu + i;
270 int tmp = p->b->core.qual < d->c->cap_mapQ? p->b->core.qual : d->c->cap_mapQ;
271 rms_aux += tmp * tmp;
272 pileup_seq(p, pos, d->len, d->ref);
275 rms_aux = (uint64_t)(sqrt((double)rms_aux / n) + .499);
276 if (rms_mapq < 0) rms_mapq = rms_aux;
279 for (i = 0; i < n; ++i) {
280 const bam_pileup1_t *p = pu + i;
281 int c = bam1_qual(p->b)[p->qpos] + 33;
282 if (c > 126) c = 126;
285 if (d->format & BAM_PLF_2ND) { // print 2nd calls and qualities
286 const unsigned char *q;
288 for (i = 0; i < n; ++i) {
289 const bam_pileup1_t *p = pu + i;
290 q = bam_aux_get(p->b, "E2");
291 putchar(q? q[p->qpos + 1] : 'N');
294 for (i = 0; i < n; ++i) {
295 const bam_pileup1_t *p = pu + i;
296 q = bam_aux_get(p->b, "U2");
297 putchar(q? q[p->qpos + 1] : '!');
300 // print mapping quality if -s is flagged on the command line
301 if (d->format & BAM_PLF_SIMPLE) {
303 for (i = 0; i < n; ++i) {
304 int c = pu[i].b->core.qual + 33;
305 if (c > 126) c = 126;
309 // print read position
310 if (d->format & BAM_PLF_READPOS) {
312 for (i = 0; i < n; ++i) {
314 int l = pu[i].b->core.l_qseq;
315 printf("%d,", x < l/2? x+1 : -((l-1)-x+1));
319 // print the indel line if r has been calculated. This only happens if:
320 // a) -c or -i are flagged, AND b) the reference sequence is available
322 printf("%s\t%d\t*\t", d->h->target_name[tid], pos + 1);
323 if (r->gt < 2) printf("%s/%s\t", r->s[r->gt], r->s[r->gt]);
324 else printf("%s/%s\t", r->s[0], r->s[1]);
325 printf("%d\t%d\t", r->q_cns, r->q_ref);
326 printf("%d\t%d\t", rms_mapq, n);
327 printf("%s\t%s\t", r->s[0], r->s[1]);
328 //printf("%d\t%d\t", r->gl[0], r->gl[1]);
329 printf("%d\t%d\t%d\t", r->cnt1, r->cnt2, r->cnt_anti);
330 printf("%d\t%d\n", r->cnt_ref, r->cnt_ambi);
331 bam_maqindel_ret_destroy(r);
336 int bam_pileup(int argc, char *argv[])
339 char *fn_list = 0, *fn_fa = 0, *fn_pos = 0;
340 pu_data_t *d = (pu_data_t*)calloc(1, sizeof(pu_data_t));
342 d->tid = -1; d->mask = BAM_DEF_MASK;
343 d->c = bam_maqcns_init();
344 d->ido = bam_maqindel_opt_init();
345 while ((c = getopt(argc, argv, "st:f:cT:N:r:l:d:im:gI:G:vM:S2aR:P")) >= 0) {
347 case 'a': d->c->is_soap = 1; break;
348 case 's': d->format |= BAM_PLF_SIMPLE; break;
349 case 't': fn_list = strdup(optarg); break;
350 case 'l': fn_pos = strdup(optarg); break;
351 case 'f': fn_fa = strdup(optarg); break;
352 case 'T': d->c->theta = atof(optarg); break;
353 case 'N': d->c->n_hap = atoi(optarg); break;
354 case 'r': d->c->het_rate = atof(optarg); break;
355 case 'M': d->c->cap_mapQ = atoi(optarg); break;
356 case 'd': d->max_depth = atoi(optarg); break;
357 case 'c': d->format |= BAM_PLF_CNS; break;
358 case 'i': d->format |= BAM_PLF_INDEL_ONLY; break;
359 case 'v': d->format |= BAM_PLF_VAR_ONLY; break;
360 case 'm': d->mask = strtol(optarg, 0, 0); break;
361 case 'g': d->format |= BAM_PLF_GLF; break;
362 case '2': d->format |= BAM_PLF_2ND; break;
363 case 'P': d->format |= BAM_PLF_READPOS; break;
364 case 'I': d->ido->q_indel = atoi(optarg); break;
365 case 'G': d->ido->r_indel = atof(optarg); break;
366 case 'S': is_SAM = 1; break;
368 if (strcmp(optarg, "random") == 0) d->format |= BAM_PLF_RANBASE;
369 else if (strcmp(optarg, "first") == 0) d->format |= BAM_PLF_1STBASE;
370 else if (strcmp(optarg, "all") == 0) d->format |= BAM_PLF_ALLBASE;
371 else fprintf(stderr, "[bam_pileup] unrecognized -R\n");
373 default: fprintf(stderr, "Unrecognizd option '-%c'.\n", c); return 1;
376 if (fn_list) is_SAM = 1;
377 if (optind == argc) {
378 fprintf(stderr, "\n");
379 fprintf(stderr, "Usage: samtools pileup [options] <in.bam>|<in.sam>\n\n");
380 fprintf(stderr, "Option: -s simple (yet incomplete) pileup format\n");
381 fprintf(stderr, " -S the input is in SAM\n");
382 fprintf(stderr, " -a use the SOAPsnp model for SNP calling\n");
383 fprintf(stderr, " -2 output the 2nd best call and quality\n");
384 fprintf(stderr, " -i only show lines/consensus with indels\n");
385 fprintf(stderr, " -m INT filtering reads with bits in INT [%d]\n", d->mask);
386 fprintf(stderr, " -M INT cap mapping quality at INT [%d]\n", d->c->cap_mapQ);
387 fprintf(stderr, " -d INT limit maximum depth for indels [unlimited]\n");
388 fprintf(stderr, " -t FILE list of reference sequences (force -S)\n");
389 fprintf(stderr, " -l FILE list of sites at which pileup is output\n");
390 fprintf(stderr, " -f FILE reference sequence in the FASTA format\n\n");
391 fprintf(stderr, " -c output the maq consensus sequence\n");
392 fprintf(stderr, " -v print variants only (for -c)\n");
393 fprintf(stderr, " -g output in the GLFv3 format (suppressing -c/-i/-s)\n");
394 fprintf(stderr, " -T FLOAT theta in maq consensus calling model (for -c/-g) [%f]\n", d->c->theta);
395 fprintf(stderr, " -N INT number of haplotypes in the sample (for -c/-g) [%d]\n", d->c->n_hap);
396 fprintf(stderr, " -r FLOAT prior of a difference between two haplotypes (for -c/-g) [%f]\n", d->c->het_rate);
397 fprintf(stderr, " -G FLOAT prior of an indel between two haplotypes (for -c/-g) [%f]\n", d->ido->r_indel);
398 fprintf(stderr, " -I INT phred prob. of an indel in sequencing/prep. (for -c/-g) [%d]\n", d->ido->q_indel);
399 fprintf(stderr, "\n");
400 free(fn_list); free(fn_fa); free(d);
403 if (d->format & (BAM_PLF_RANBASE|BAM_PLF_1STBASE|BAM_PLF_ALLBASE)) d->format |= BAM_PLF_CNS;
404 if (fn_fa) d->fai = fai_load(fn_fa);
405 if (d->format & (BAM_PLF_CNS|BAM_PLF_GLF)) bam_maqcns_prepare(d->c); // consensus calling
406 if (d->format & BAM_PLF_GLF) { // for glf output
408 h = glf3_header_init();
409 d->fp_glf = bgzf_fdopen(fileno(stdout), "w");
410 glf3_header_write(d->fp_glf, h);
411 glf3_header_destroy(h);
413 if (d->fai == 0 && (d->format & (BAM_PLF_CNS|BAM_PLF_INDEL_ONLY)))
414 fprintf(stderr, "[bam_pileup] indels will not be called when -f is absent.\n");
415 if (fn_fa && is_SAM && fn_list == 0) fn_list = samfaipath(fn_fa);
419 fp = is_SAM? samopen(argv[optind], "r", fn_list) : samopen(argv[optind], "rb", 0);
420 if (fp == 0 || fp->header == 0) {
421 fprintf(stderr, "[bam_pileup] fail to read the header: non-exisiting file or wrong format.\n");
425 if (fn_pos) d->hash = load_pos(fn_pos, d->h);
426 sampileup(fp, d->mask, pileup_func, d);
427 samclose(fp); // d->h will be destroyed here
431 if (d->format & BAM_PLF_GLF) bgzf_close(d->fp_glf);
432 if (fn_pos) { // free the hash table
434 for (k = kh_begin(d->hash); k < kh_end(d->hash); ++k)
435 if (kh_exist(d->hash, k)) free(kh_val(d->hash, k));
436 kh_destroy(64, d->hash);
438 free(fn_pos); free(fn_list); free(fn_fa);
439 if (d->fai) fai_destroy(d->fai);
440 bam_maqcns_destroy(d->c);
441 free(d->ido); free(d->ref); free(d);
459 static int mplp_func(void *data, bam1_t *b)
461 mplp_aux_t *ma = (mplp_aux_t*)data;
462 if (ma->iter) return bam_iter_read(ma->fp, ma->iter, b);
463 return bam_read1(ma->fp, b);
466 static int mpileup(mplp_conf_t *conf, int n, char **fn)
469 int i, tid, pos, *n_plp, beg0 = 0, end0 = 1u<<29, ref_len, ref_tid;
470 const bam_pileup1_t **plp;
475 data = calloc(n, sizeof(void*));
476 plp = calloc(n, sizeof(void*));
477 n_plp = calloc(n, sizeof(int*));
478 // read the header and initialize data
479 for (i = 0; i < n; ++i) {
481 data[i] = calloc(1, sizeof(mplp_aux_t));
482 data[i]->fp = bam_open(fn[i], "r");
483 h_tmp = bam_header_read(data[i]->fp);
487 idx = bam_index_load(fn[i]);
489 fprintf(stderr, "[%s] fail to load index for %d-th input.\n", __func__, i+1);
492 if (bam_parse_region(h_tmp, conf->reg, &tid, &beg, &end) < 0) {
493 fprintf(stderr, "[%s] malformatted region or wrong seqname for %d-th input.\n", __func__, i+1);
496 if (i == 0) beg0 = beg, end0 = end;
497 data[i]->iter = bam_iter_query(idx, tid, beg, end);
498 bam_index_destroy(idx);
500 if (i == 0) h = h_tmp;
502 // FIXME: to check consistency
503 bam_header_destroy(h_tmp);
507 ref_tid = -1; ref = 0;
508 iter = bam_mplp_init(n, mplp_func, (void**)data);
509 while (bam_mplp_auto(iter, &tid, &pos, n_plp, plp) > 0) {
510 if (conf->reg && (pos < beg0 || pos >= end0)) continue; // out of the region requested
511 if (tid != ref_tid) {
513 if (conf->fai) ref = fai_fetch(conf->fai, h->target_name[tid], &ref_len);
516 printf("%s\t%d\t%c", h->target_name[tid], pos + 1, (ref && pos < ref_len)? ref[pos] : 'N');
517 for (i = 0; i < n; ++i) {
519 printf("\t%d\t", n_plp[i]);
520 if (n_plp[i] == 0) printf("*\t*");
522 for (j = 0; j < n_plp[i]; ++j)
523 pileup_seq(plp[i] + j, pos, ref_len, ref);
525 for (j = 0; j < n_plp[i]; ++j) {
526 const bam_pileup1_t *p = plp[i] + j;
527 int c = bam1_qual(p->b)[p->qpos] + 33;
528 if (c > 126) c = 126;
535 bam_mplp_destroy(iter);
536 bam_header_destroy(h);
537 for (i = 0; i < n; ++i) {
538 bam_close(data[i]->fp);
539 if (data[i]->iter) bam_iter_destroy(data[i]->iter);
542 free(data); free(plp); free(ref); free(n_plp);
546 int bam_mpileup(int argc, char *argv[])
550 memset(&mplp, 0, sizeof(mplp_conf_t));
551 while ((c = getopt(argc, argv, "f:r:")) >= 0) {
554 mplp.fai = fai_load(optarg);
555 if (mplp.fai == 0) return 1;
557 case 'r': mplp.reg = strdup(optarg);
561 fprintf(stderr, "Usage: samtools mpileup [-r reg] [-f in.fa] in1.bam [in2.bam [...]]\n");
564 mpileup(&mplp, argc - optind, argv + optind);
566 if (mplp.fai) fai_destroy(mplp.fai);