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
28 bam_maqindel_opt_t *ido;
32 int tid, len, last_pos;
34 int max_depth; // for indel calling, ignore reads with the depth too high. 0 for unlimited
36 glfFile fp_glf; // for glf output only
39 char **__bam_get_lines(const char *fn, int *_n);
40 void bam_init_header_hash(bam_header_t *header);
41 int32_t bam_get_tid(const bam_header_t *header, const char *seq_name);
43 static khash_t(64) *load_pos(const char *fn, bam_header_t *h)
46 int i, j, n, *fields, max_fields;
48 bam_init_header_hash(h);
49 list = __bam_get_lines(fn, &n);
51 max_fields = 0; fields = 0;
52 for (i = 0; i < n; ++i) {
54 int chr, n_fields, ret;
57 n_fields = ksplit_core(str, 0, &max_fields, &fields);
58 if (n_fields < 2) continue;
59 chr = bam_get_tid(h, str + fields[0]);
61 fprintf(stderr, "[load_pos] unknown reference sequence name: %s\n", str + fields[0]);
64 x = (uint64_t)chr << 32 | (atoi(str + fields[1]) - 1);
65 k = kh_put(64, hash, x, &ret);
67 fprintf(stderr, "[load_pos] position %s:%s has been loaded.\n", str+fields[0], str+fields[1]);
73 for (j = 2; j < n_fields; ++j) {
74 char *s = str + fields[j];
75 if ((*s != '+' && *s != '-') || !isdigit(s[1])) break;
77 if (j > 2) { // update kh_val()
79 q = kh_val(hash, k) = (int*)calloc(j - 1, sizeof(int));
80 q[0] = j - 2; z = j; y = 1;
81 for (j = 2; j < z; ++j)
82 q[y++] = atoi(str + fields[j]);
87 free(list); free(fields);
91 // an analogy to pileup_func() below
92 static int glt3_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *pu, void *data)
94 pu_data_t *d = (pu_data_t*)data;
95 bam_maqindel_ret_t *r = 0;
96 int rb, *proposed_indels = 0;
101 fprintf(stderr, "[glt3_func] reference sequence is required for generating GLT. Abort!\n");
104 if (d->hash) { // only output a list of sites
105 khint_t k = kh_get(64, d->hash, (uint64_t)tid<<32|pos);
106 if (k == kh_end(d->hash)) return 0;
107 proposed_indels = kh_val(d->hash, k);
110 if (d->fai && (int)tid != d->tid) {
111 if (d->ref) { // then write the end mark
112 g3->rtype = GLF3_RTYPE_END;
113 glf3_write1(d->fp_glf, g3);
115 glf3_ref_write(d->fp_glf, d->h->target_name[tid], d->h->target_len[tid]); // write reference
117 d->ref = fai_fetch(d->fai, d->h->target_name[tid], &d->len);
121 rb = (d->ref && (int)pos < d->len)? d->ref[pos] : 'N';
122 g = bam_maqcns_glfgen(n, pu, bam_nt16_table[rb], d->c);
123 memcpy(g3, g, sizeof(glf1_t));
124 g3->rtype = GLF3_RTYPE_SUB;
125 g3->offset = pos - d->last_pos;
127 glf3_write1(d->fp_glf, g3);
129 int m = (!d->max_depth || d->max_depth>n) ? n : d->max_depth;
131 r = bam_maqindel(m, pos, d->ido, pu, d->ref, proposed_indels[0], proposed_indels+1);
132 else r = bam_maqindel(m, pos, d->ido, pu, d->ref, 0, 0);
134 if (r) { // then write indel line
135 int het = 3 * n, min;
137 if (min > r->gl[0]) min = r->gl[0];
138 if (min > r->gl[1]) min = r->gl[1];
140 g3->rtype = GLF3_RTYPE_INDEL;
141 memset(g3->lk, 0, 10);
142 g3->lk[0] = r->gl[0] - min < 255? r->gl[0] - min : 255;
143 g3->lk[1] = r->gl[1] - min < 255? r->gl[1] - min : 255;
144 g3->lk[2] = het - min < 255? het - min : 255;
146 g3->indel_len[0] = r->indel1;
147 g3->indel_len[1] = r->indel2;
148 g3->min_lk = min < 255? min : 255;
149 g3->max_len = (abs(r->indel1) > abs(r->indel2)? abs(r->indel1) : abs(r->indel2)) + 1;
150 g3->indel_seq[0] = strdup(r->s[0]+1);
151 g3->indel_seq[1] = strdup(r->s[1]+1);
152 glf3_write1(d->fp_glf, g3);
153 bam_maqindel_ret_destroy(r);
160 static int pileup_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *pu, void *data)
162 pu_data_t *d = (pu_data_t*)data;
163 bam_maqindel_ret_t *r = 0;
164 int i, j, rb, rms_mapq = -1, *proposed_indels = 0;
168 // if GLF is required, suppress -c completely
169 if (d->format & BAM_PLF_GLF) return glt3_func(tid, pos, n, pu, data);
170 // if d->hash is initialized, only output the sites in the hash table
172 khint_t k = kh_get(64, d->hash, (uint64_t)tid<<32|pos);
173 if (k == kh_end(d->hash)) return 0;
174 proposed_indels = kh_val(d->hash, k);
176 // update d->ref if necessary
177 if (d->fai && (int)tid != d->tid) {
179 d->ref = faidx_fetch_seq(d->fai, d->h->target_name[tid], 0, 0x7fffffff, &d->len);
182 rb = (d->ref && (int)pos < d->len)? d->ref[pos] : 'N';
183 // when the indel-only mode is asked for, return if no reads mapped with indels
184 if (d->format & BAM_PLF_INDEL_ONLY) {
185 for (i = 0; i < n; ++i)
186 if (pu[i].indel != 0) break;
187 if (i == n) return 0;
189 // call the consensus and indel
190 if (d->format & BAM_PLF_CNS) { // call consensus
191 if (d->format & (BAM_PLF_RANBASE|BAM_PLF_1STBASE)) { // use a random base or the 1st base as the consensus call
192 const bam_pileup1_t *p = (d->format & BAM_PLF_1STBASE)? pu : pu + (int)(drand48() * n);
193 int q = bam1_qual(p->b)[p->qpos];
194 int mapQ = p->b->core.qual < d->c->cap_mapQ? p->b->core.qual : d->c->cap_mapQ;
195 uint32_t b = bam1_seqi(bam1_seq(p->b), p->qpos);
196 cns = b<<28 | 0xf<<24 | mapQ<<16 | q<<8;
197 } else if (d->format & BAM_PLF_ALLBASE) { // collapse all bases
200 for (i = 0; i < n; ++i) {
201 const bam_pileup1_t *p = pu + i;
202 int q = p->b->core.qual < d->c->cap_mapQ? p->b->core.qual : d->c->cap_mapQ;
203 b |= bam1_seqi(bam1_seq(p->b), p->qpos);
206 rmsQ = (uint64_t)(sqrt((double)rmsQ / n) + .499);
207 cns = b<<28 | 0xf<<24 | rmsQ<<16 | 60<<8;
208 } else cns = bam_maqcns_call(n, pu, d->c);
210 if ((d->format & (BAM_PLF_CNS|BAM_PLF_INDEL_ONLY)) && d->ref && pos < d->len) { // call indels
211 int m = (!d->max_depth || d->max_depth>n) ? n : d->max_depth;
212 if (proposed_indels) // the first element gives the size of the array
213 r = bam_maqindel(m, pos, d->ido, pu, d->ref, proposed_indels[0], proposed_indels+1);
214 else r = bam_maqindel(m, pos, d->ido, pu, d->ref, 0, 0);
216 // when only variant sites are asked for, test if the site is a variant
217 if ((d->format & BAM_PLF_CNS) && (d->format & BAM_PLF_VAR_ONLY)) {
218 if (!(bam_nt16_table[rb] != 15 && cns>>28 != bam_nt16_table[rb])) { // not a SNP
219 if (!(r && (r->gt == 2 || strcmp(r->s[r->gt], "*")))) { // not an indel
220 if (r) bam_maqindel_ret_destroy(r);
225 // print the first 3 columns
226 printf("%s\t%d\t%c\t", d->h->target_name[tid], pos + 1, rb);
227 // print consensus information if required
228 if (d->format & BAM_PLF_CNS) {
229 int ref_q, rb4 = bam_nt16_table[rb];
231 if (rb4 != 15 && cns>>28 != 15 && cns>>28 != rb4) { // a SNP
232 ref_q = ((cns>>24&0xf) == rb4)? cns>>8&0xff : (cns>>8&0xff) + (cns&0xff);
233 if (ref_q > 255) ref_q = 255;
235 rms_mapq = cns>>16&0xff;
236 printf("%c\t%d\t%d\t%d\t", bam_nt16_rev_table[cns>>28], cns>>8&0xff, ref_q, rms_mapq);
238 // print pileup sequences
240 rms_aux = 0; // we need to recalculate rms_mapq when -c is not flagged on the command line
241 for (i = 0; i < n; ++i) {
242 const bam_pileup1_t *p = pu + i;
243 int tmp = p->b->core.qual < d->c->cap_mapQ? p->b->core.qual : d->c->cap_mapQ;
244 rms_aux += tmp * tmp;
245 if (p->is_head) printf("^%c", p->b->core.qual > 93? 126 : p->b->core.qual + 33);
247 int c = bam_nt16_rev_table[bam1_seqi(bam1_seq(p->b), p->qpos)];
248 if (c == '=' || toupper(c) == toupper(rb)) c = bam1_strand(p->b)? ',' : '.';
249 else c = bam1_strand(p->b)? tolower(c) : toupper(c);
252 printf("+%d", p->indel);
253 for (j = 1; j <= p->indel; ++j) {
254 c = bam_nt16_rev_table[bam1_seqi(bam1_seq(p->b), p->qpos + j)];
255 putchar(bam1_strand(p->b)? tolower(c) : toupper(c));
257 } else if (p->indel < 0) {
258 printf("%d", p->indel);
259 for (j = 1; j <= -p->indel; ++j) {
260 c = (d->ref && (int)pos+j < d->len)? d->ref[pos+j] : 'N';
261 putchar(bam1_strand(p->b)? tolower(c) : toupper(c));
265 if (p->is_tail) putchar('$');
268 rms_aux = (uint64_t)(sqrt((double)rms_aux / n) + .499);
269 if (rms_mapq < 0) rms_mapq = rms_aux;
272 for (i = 0; i < n; ++i) {
273 const bam_pileup1_t *p = pu + i;
274 int c = bam1_qual(p->b)[p->qpos] + 33;
275 if (c > 126) c = 126;
278 if (d->format & BAM_PLF_2ND) { // print 2nd calls and qualities
279 const unsigned char *q;
281 for (i = 0; i < n; ++i) {
282 const bam_pileup1_t *p = pu + i;
283 q = bam_aux_get(p->b, "E2");
284 putchar(q? q[p->qpos + 1] : 'N');
287 for (i = 0; i < n; ++i) {
288 const bam_pileup1_t *p = pu + i;
289 q = bam_aux_get(p->b, "U2");
290 putchar(q? q[p->qpos + 1] : '!');
293 // print mapping quality if -s is flagged on the command line
294 if (d->format & BAM_PLF_SIMPLE) {
296 for (i = 0; i < n; ++i) {
297 int c = pu[i].b->core.qual + 33;
298 if (c > 126) c = 126;
303 // print the indel line if r has been calculated. This only happens if:
304 // a) -c or -i are flagged, AND b) the reference sequence is available
306 printf("%s\t%d\t*\t", d->h->target_name[tid], pos + 1);
307 if (r->gt < 2) printf("%s/%s\t", r->s[r->gt], r->s[r->gt]);
308 else printf("%s/%s\t", r->s[0], r->s[1]);
309 printf("%d\t%d\t", r->q_cns, r->q_ref);
310 printf("%d\t%d\t", rms_mapq, n);
311 printf("%s\t%s\t", r->s[0], r->s[1]);
312 //printf("%d\t%d\t", r->gl[0], r->gl[1]);
313 printf("%d\t%d\t%d\t", r->cnt1, r->cnt2, r->cnt_anti);
314 printf("%d\t%d\n", r->cnt_ref, r->cnt_ambi);
315 bam_maqindel_ret_destroy(r);
320 int bam_pileup(int argc, char *argv[])
323 char *fn_list = 0, *fn_fa = 0, *fn_pos = 0;
324 pu_data_t *d = (pu_data_t*)calloc(1, sizeof(pu_data_t));
326 d->tid = -1; d->mask = BAM_DEF_MASK;
327 d->c = bam_maqcns_init();
328 d->ido = bam_maqindel_opt_init();
329 while ((c = getopt(argc, argv, "st:f:cT:N:r:l:d:im:gI:G:vM:S2aR:")) >= 0) {
331 case 'a': d->c->is_soap = 1; break;
332 case 's': d->format |= BAM_PLF_SIMPLE; break;
333 case 't': fn_list = strdup(optarg); break;
334 case 'l': fn_pos = strdup(optarg); break;
335 case 'f': fn_fa = strdup(optarg); break;
336 case 'T': d->c->theta = atof(optarg); break;
337 case 'N': d->c->n_hap = atoi(optarg); break;
338 case 'r': d->c->het_rate = atof(optarg); break;
339 case 'M': d->c->cap_mapQ = atoi(optarg); break;
340 case 'd': d->max_depth = atoi(optarg); break;
341 case 'c': d->format |= BAM_PLF_CNS; break;
342 case 'i': d->format |= BAM_PLF_INDEL_ONLY; break;
343 case 'v': d->format |= BAM_PLF_VAR_ONLY; break;
344 case 'm': d->mask = strtol(optarg, 0, 0); break;
345 case 'g': d->format |= BAM_PLF_GLF; break;
346 case '2': d->format |= BAM_PLF_2ND; break;
347 case 'I': d->ido->q_indel = atoi(optarg); break;
348 case 'G': d->ido->r_indel = atof(optarg); break;
349 case 'S': is_SAM = 1; break;
351 if (strcmp(optarg, "random") == 0) d->format |= BAM_PLF_RANBASE;
352 else if (strcmp(optarg, "first") == 0) d->format |= BAM_PLF_1STBASE;
353 else if (strcmp(optarg, "all") == 0) d->format |= BAM_PLF_ALLBASE;
354 else fprintf(stderr, "[bam_pileup] unrecognized -R\n");
356 default: fprintf(stderr, "Unrecognizd option '-%c'.\n", c); return 1;
359 if (fn_list) is_SAM = 1;
360 if (optind == argc) {
361 fprintf(stderr, "\n");
362 fprintf(stderr, "Usage: samtools pileup [options] <in.bam>|<in.sam>\n\n");
363 fprintf(stderr, "Option: -s simple (yet incomplete) pileup format\n");
364 fprintf(stderr, " -S the input is in SAM\n");
365 fprintf(stderr, " -a use the SOAPsnp model for SNP calling\n");
366 fprintf(stderr, " -2 output the 2nd best call and quality\n");
367 fprintf(stderr, " -i only show lines/consensus with indels\n");
368 fprintf(stderr, " -m INT filtering reads with bits in INT [%d]\n", d->mask);
369 fprintf(stderr, " -M INT cap mapping quality at INT [%d]\n", d->c->cap_mapQ);
370 fprintf(stderr, " -d INT limit maximum depth for indels [unlimited]\n");
371 fprintf(stderr, " -t FILE list of reference sequences (force -S)\n");
372 fprintf(stderr, " -l FILE list of sites at which pileup is output\n");
373 fprintf(stderr, " -f FILE reference sequence in the FASTA format\n\n");
374 fprintf(stderr, " -c output the maq consensus sequence\n");
375 fprintf(stderr, " -v print variants only (for -c)\n");
376 fprintf(stderr, " -g output in the GLFv3 format (suppressing -c/-i/-s)\n");
377 fprintf(stderr, " -T FLOAT theta in maq consensus calling model (for -c/-g) [%f]\n", d->c->theta);
378 fprintf(stderr, " -N INT number of haplotypes in the sample (for -c/-g) [%d]\n", d->c->n_hap);
379 fprintf(stderr, " -r FLOAT prior of a difference between two haplotypes (for -c/-g) [%f]\n", d->c->het_rate);
380 fprintf(stderr, " -G FLOAT prior of an indel between two haplotypes (for -c/-g) [%f]\n", d->ido->r_indel);
381 fprintf(stderr, " -I INT phred prob. of an indel in sequencing/prep. (for -c/-g) [%d]\n", d->ido->q_indel);
382 fprintf(stderr, "\n");
383 free(fn_list); free(fn_fa); free(d);
386 if (d->format & (BAM_PLF_RANBASE|BAM_PLF_1STBASE|BAM_PLF_ALLBASE)) d->format |= BAM_PLF_CNS;
387 if (fn_fa) d->fai = fai_load(fn_fa);
388 if (d->format & (BAM_PLF_CNS|BAM_PLF_GLF)) bam_maqcns_prepare(d->c); // consensus calling
389 if (d->format & BAM_PLF_GLF) { // for glf output
391 h = glf3_header_init();
392 d->fp_glf = bgzf_fdopen(fileno(stdout), "w");
393 glf3_header_write(d->fp_glf, h);
394 glf3_header_destroy(h);
396 if (d->fai == 0 && (d->format & (BAM_PLF_CNS|BAM_PLF_INDEL_ONLY)))
397 fprintf(stderr, "[bam_pileup] indels will not be called when -f is absent.\n");
398 if (fn_fa && is_SAM && fn_list == 0) fn_list = samfaipath(fn_fa);
402 fp = is_SAM? samopen(argv[optind], "r", fn_list) : samopen(argv[optind], "rb", 0);
403 if (fp == 0 || fp->header == 0) {
404 fprintf(stderr, "[bam_pileup] fail to read the header: non-exisiting file or wrong format.\n");
408 if (fn_pos) d->hash = load_pos(fn_pos, d->h);
409 sampileup(fp, d->mask, pileup_func, d);
410 samclose(fp); // d->h will be destroyed here
414 if (d->format & BAM_PLF_GLF) bgzf_close(d->fp_glf);
415 if (fn_pos) { // free the hash table
417 for (k = kh_begin(d->hash); k < kh_end(d->hash); ++k)
418 if (kh_exist(d->hash, k)) free(kh_val(d->hash, k));
419 kh_destroy(64, d->hash);
421 free(fn_pos); free(fn_list); free(fn_fa);
422 if (d->fai) fai_destroy(d->fai);
423 bam_maqcns_destroy(d->c);
424 free(d->ido); free(d->ref); free(d);