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;
35 glfFile fp_glf; // for glf output only
38 char **__bam_get_lines(const char *fn, int *_n);
39 void bam_init_header_hash(bam_header_t *header);
40 int32_t bam_get_tid(const bam_header_t *header, const char *seq_name);
42 static khash_t(64) *load_pos(const char *fn, bam_header_t *h)
45 int i, j, n, *fields, max_fields;
47 bam_init_header_hash(h);
48 list = __bam_get_lines(fn, &n);
50 max_fields = 0; fields = 0;
51 for (i = 0; i < n; ++i) {
53 int chr, n_fields, ret;
56 n_fields = ksplit_core(str, 0, &max_fields, &fields);
57 if (n_fields < 2) continue;
58 chr = bam_get_tid(h, str + fields[0]);
60 fprintf(stderr, "[load_pos] unknown reference sequence name: %s\n", str + fields[0]);
63 x = (uint64_t)chr << 32 | (atoi(str + fields[1]) - 1);
64 k = kh_put(64, hash, x, &ret);
66 fprintf(stderr, "[load_pos] position %s:%s has been loaded.\n", str+fields[0], str+fields[1]);
72 for (j = 2; j < n_fields; ++j) {
73 char *s = str + fields[j];
74 if ((*s != '+' && *s != '-') || !isdigit(s[1])) break;
76 if (j > 2) { // update kh_val()
78 q = kh_val(hash, k) = (int*)calloc(j - 1, sizeof(int));
79 q[0] = j - 2; z = j; y = 1;
80 for (j = 2; j < z; ++j)
81 q[y++] = atoi(str + fields[j]);
86 free(list); free(fields);
90 // an analogy to pileup_func() below
91 static int glt3_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *pu, void *data)
93 pu_data_t *d = (pu_data_t*)data;
94 bam_maqindel_ret_t *r = 0;
95 int rb, *proposed_indels = 0;
100 fprintf(stderr, "[glt3_func] reference sequence is required for generating GLT. Abort!\n");
103 if (d->hash) { // only output a list of sites
104 khint_t k = kh_get(64, d->hash, (uint64_t)tid<<32|pos);
105 if (k == kh_end(d->hash)) return 0;
106 proposed_indels = kh_val(d->hash, k);
109 if (d->fai && (int)tid != d->tid) {
110 if (d->ref) { // then write the end mark
111 g3->rtype = GLF3_RTYPE_END;
112 glf3_write1(d->fp_glf, g3);
114 glf3_ref_write(d->fp_glf, d->h->target_name[tid], d->h->target_len[tid]); // write reference
116 d->ref = fai_fetch(d->fai, d->h->target_name[tid], &d->len);
120 rb = (d->ref && (int)pos < d->len)? d->ref[pos] : 'N';
121 g = bam_maqcns_glfgen(n, pu, bam_nt16_table[rb], d->c);
122 memcpy(g3, g, sizeof(glf1_t));
123 g3->rtype = GLF3_RTYPE_SUB;
124 g3->offset = pos - d->last_pos;
126 glf3_write1(d->fp_glf, g3);
129 r = bam_maqindel(n, pos, d->ido, pu, d->ref, proposed_indels[0], proposed_indels+1);
130 else r = bam_maqindel(n, pos, d->ido, pu, d->ref, 0, 0);
132 if (r) { // then write indel line
133 int het = 3 * n, min;
135 if (min > r->gl[0]) min = r->gl[0];
136 if (min > r->gl[1]) min = r->gl[1];
138 g3->rtype = GLF3_RTYPE_INDEL;
139 memset(g3->lk, 0, 10);
140 g3->lk[0] = r->gl[0] - min < 255? r->gl[0] - min : 255;
141 g3->lk[1] = r->gl[1] - min < 255? r->gl[1] - min : 255;
142 g3->lk[2] = het - min < 255? het - min : 255;
144 g3->indel_len[0] = r->indel1;
145 g3->indel_len[1] = r->indel2;
146 g3->min_lk = min < 255? min : 255;
147 g3->max_len = (abs(r->indel1) > abs(r->indel2)? abs(r->indel1) : abs(r->indel2)) + 1;
148 g3->indel_seq[0] = strdup(r->s[0]+1);
149 g3->indel_seq[1] = strdup(r->s[1]+1);
150 glf3_write1(d->fp_glf, g3);
151 bam_maqindel_ret_destroy(r);
158 static int pileup_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *pu, void *data)
160 pu_data_t *d = (pu_data_t*)data;
161 bam_maqindel_ret_t *r = 0;
162 int i, j, rb, rms_mapq = -1, *proposed_indels = 0;
166 // if GLF is required, suppress -c completely
167 if (d->format & BAM_PLF_GLF) return glt3_func(tid, pos, n, pu, data);
168 // if d->hash is initialized, only output the sites in the hash table
170 khint_t k = kh_get(64, d->hash, (uint64_t)tid<<32|pos);
171 if (k == kh_end(d->hash)) return 0;
172 proposed_indels = kh_val(d->hash, k);
174 // update d->ref if necessary
175 if (d->fai && (int)tid != d->tid) {
177 d->ref = faidx_fetch_seq(d->fai, d->h->target_name[tid], 0, 0x7fffffff, &d->len);
180 rb = (d->ref && (int)pos < d->len)? d->ref[pos] : 'N';
181 // when the indel-only mode is asked for, return if no reads mapped with indels
182 if (d->format & BAM_PLF_INDEL_ONLY) {
183 for (i = 0; i < n; ++i)
184 if (pu[i].indel != 0) break;
185 if (i == n) return 0;
187 // call the consensus and indel
188 if (d->format & BAM_PLF_CNS) { // call consensus
189 if (d->format & (BAM_PLF_RANBASE|BAM_PLF_1STBASE)) { // use a random base or the 1st base as the consensus call
190 const bam_pileup1_t *p = (d->format & BAM_PLF_1STBASE)? pu : pu + (int)(drand48() * n);
191 int q = bam1_qual(p->b)[p->qpos];
192 int mapQ = p->b->core.qual < d->c->cap_mapQ? p->b->core.qual : d->c->cap_mapQ;
193 uint32_t b = bam1_seqi(bam1_seq(p->b), p->qpos);
194 cns = b<<28 | 0xf<<24 | mapQ<<16 | q<<8;
195 } else if (d->format & BAM_PLF_ALLBASE) { // collapse all bases
198 for (i = 0; i < n; ++i) {
199 const bam_pileup1_t *p = pu + i;
200 int q = p->b->core.qual < d->c->cap_mapQ? p->b->core.qual : d->c->cap_mapQ;
201 b |= bam1_seqi(bam1_seq(p->b), p->qpos);
204 rmsQ = (uint64_t)(sqrt((double)rmsQ / n) + .499);
205 cns = b<<28 | 0xf<<24 | rmsQ<<16 | 60<<8;
206 } else cns = bam_maqcns_call(n, pu, d->c);
208 if ((d->format & (BAM_PLF_CNS|BAM_PLF_INDEL_ONLY)) && d->ref && pos < d->len) { // call indels
209 if (proposed_indels) // the first element gives the size of the array
210 r = bam_maqindel(n, pos, d->ido, pu, d->ref, proposed_indels[0], proposed_indels+1);
211 else r = bam_maqindel(n, pos, d->ido, pu, d->ref, 0, 0);
213 // when only variant sites are asked for, test if the site is a variant
214 if ((d->format & BAM_PLF_CNS) && (d->format & BAM_PLF_VAR_ONLY)) {
215 if (!(bam_nt16_table[rb] != 15 && cns>>28 != bam_nt16_table[rb])) { // not a SNP
216 if (!(r && (r->gt == 2 || strcmp(r->s[r->gt], "*")))) { // not an indel
217 if (r) bam_maqindel_ret_destroy(r);
222 // print the first 3 columns
223 printf("%s\t%d\t%c\t", d->h->target_name[tid], pos + 1, rb);
224 // print consensus information if required
225 if (d->format & BAM_PLF_CNS) {
226 int ref_q, rb4 = bam_nt16_table[rb];
228 if (rb4 != 15 && cns>>28 != 15 && cns>>28 != rb4) { // a SNP
229 ref_q = ((cns>>24&0xf) == rb4)? cns>>8&0xff : (cns>>8&0xff) + (cns&0xff);
230 if (ref_q > 255) ref_q = 255;
232 rms_mapq = cns>>16&0xff;
233 printf("%c\t%d\t%d\t%d\t", bam_nt16_rev_table[cns>>28], cns>>8&0xff, ref_q, rms_mapq);
235 // print pileup sequences
237 rms_aux = 0; // we need to recalculate rms_mapq when -c is not flagged on the command line
238 for (i = 0; i < n; ++i) {
239 const bam_pileup1_t *p = pu + i;
240 int tmp = p->b->core.qual < d->c->cap_mapQ? p->b->core.qual : d->c->cap_mapQ;
241 rms_aux += tmp * tmp;
242 if (p->is_head) printf("^%c", p->b->core.qual > 93? 126 : p->b->core.qual + 33);
244 int c = bam_nt16_rev_table[bam1_seqi(bam1_seq(p->b), p->qpos)];
245 if (c == '=' || toupper(c) == toupper(rb)) c = bam1_strand(p->b)? ',' : '.';
246 else c = bam1_strand(p->b)? tolower(c) : toupper(c);
249 printf("+%d", p->indel);
250 for (j = 1; j <= p->indel; ++j) {
251 c = bam_nt16_rev_table[bam1_seqi(bam1_seq(p->b), p->qpos + j)];
252 putchar(bam1_strand(p->b)? tolower(c) : toupper(c));
254 } else if (p->indel < 0) {
255 printf("%d", p->indel);
256 for (j = 1; j <= -p->indel; ++j) {
257 c = (d->ref && (int)pos+j < d->len)? d->ref[pos+j] : 'N';
258 putchar(bam1_strand(p->b)? tolower(c) : toupper(c));
262 if (p->is_tail) putchar('$');
265 rms_aux = (uint64_t)(sqrt((double)rms_aux / n) + .499);
266 if (rms_mapq < 0) rms_mapq = rms_aux;
269 for (i = 0; i < n; ++i) {
270 const bam_pileup1_t *p = pu + i;
271 int c = bam1_qual(p->b)[p->qpos] + 33;
272 if (c > 126) c = 126;
275 if (d->format & BAM_PLF_2ND) { // print 2nd calls and qualities
276 const unsigned char *q;
278 for (i = 0; i < n; ++i) {
279 const bam_pileup1_t *p = pu + i;
280 q = bam_aux_get(p->b, "E2");
281 putchar(q? q[p->qpos + 1] : 'N');
284 for (i = 0; i < n; ++i) {
285 const bam_pileup1_t *p = pu + i;
286 q = bam_aux_get(p->b, "U2");
287 putchar(q? q[p->qpos + 1] : '!');
290 // print mapping quality if -s is flagged on the command line
291 if (d->format & BAM_PLF_SIMPLE) {
293 for (i = 0; i < n; ++i) {
294 int c = pu[i].b->core.qual + 33;
295 if (c > 126) c = 126;
300 // print the indel line if r has been calculated. This only happens if:
301 // a) -c or -i are flagged, AND b) the reference sequence is available
303 printf("%s\t%d\t*\t", d->h->target_name[tid], pos + 1);
304 if (r->gt < 2) printf("%s/%s\t", r->s[r->gt], r->s[r->gt]);
305 else printf("%s/%s\t", r->s[0], r->s[1]);
306 printf("%d\t%d\t", r->q_cns, r->q_ref);
307 printf("%d\t%d\t", rms_mapq, n);
308 printf("%s\t%s\t", r->s[0], r->s[1]);
309 //printf("%d\t%d\t", r->gl[0], r->gl[1]);
310 printf("%d\t%d\t%d\t", r->cnt1, r->cnt2, r->cnt_anti);
311 printf("%d\t%d\n", r->cnt_ref, r->cnt_ambi);
312 bam_maqindel_ret_destroy(r);
317 int bam_pileup(int argc, char *argv[])
320 char *fn_list = 0, *fn_fa = 0, *fn_pos = 0;
321 pu_data_t *d = (pu_data_t*)calloc(1, sizeof(pu_data_t));
322 d->tid = -1; d->mask = BAM_DEF_MASK;
323 d->c = bam_maqcns_init();
324 d->ido = bam_maqindel_opt_init();
325 while ((c = getopt(argc, argv, "st:f:cT:N:r:l:im:gI:G:vM:S2aR:")) >= 0) {
327 case 'a': d->c->is_soap = 1; break;
328 case 's': d->format |= BAM_PLF_SIMPLE; break;
329 case 't': fn_list = strdup(optarg); break;
330 case 'l': fn_pos = strdup(optarg); break;
331 case 'f': fn_fa = strdup(optarg); break;
332 case 'T': d->c->theta = atof(optarg); break;
333 case 'N': d->c->n_hap = atoi(optarg); break;
334 case 'r': d->c->het_rate = atof(optarg); break;
335 case 'M': d->c->cap_mapQ = atoi(optarg); break;
336 case 'c': d->format |= BAM_PLF_CNS; break;
337 case 'i': d->format |= BAM_PLF_INDEL_ONLY; break;
338 case 'v': d->format |= BAM_PLF_VAR_ONLY; break;
339 case 'm': d->mask = strtol(optarg, 0, 0); break;
340 case 'g': d->format |= BAM_PLF_GLF; break;
341 case '2': d->format |= BAM_PLF_2ND; break;
342 case 'I': d->ido->q_indel = atoi(optarg); break;
343 case 'G': d->ido->r_indel = atof(optarg); break;
344 case 'S': is_SAM = 1; break;
346 if (strcmp(optarg, "random") == 0) d->format |= BAM_PLF_RANBASE;
347 else if (strcmp(optarg, "first") == 0) d->format |= BAM_PLF_1STBASE;
348 else if (strcmp(optarg, "all") == 0) d->format |= BAM_PLF_ALLBASE;
349 else fprintf(stderr, "[bam_pileup] unrecognized -R\n");
351 default: fprintf(stderr, "Unrecognizd option '-%c'.\n", c); return 1;
354 if (fn_list) is_SAM = 1;
355 if (optind == argc) {
356 fprintf(stderr, "\n");
357 fprintf(stderr, "Usage: samtools pileup [options] <in.bam>|<in.sam>\n\n");
358 fprintf(stderr, "Option: -s simple (yet incomplete) pileup format\n");
359 fprintf(stderr, " -S the input is in SAM\n");
360 fprintf(stderr, " -a use the SOAPsnp model for SNP calling\n");
361 fprintf(stderr, " -2 output the 2nd best call and quality\n");
362 fprintf(stderr, " -i only show lines/consensus with indels\n");
363 fprintf(stderr, " -m INT filtering reads with bits in INT [%d]\n", d->mask);
364 fprintf(stderr, " -M INT cap mapping quality at INT [%d]\n", d->c->cap_mapQ);
365 fprintf(stderr, " -t FILE list of reference sequences (force -S)\n");
366 fprintf(stderr, " -l FILE list of sites at which pileup is output\n");
367 fprintf(stderr, " -f FILE reference sequence in the FASTA format\n\n");
368 fprintf(stderr, " -c output the maq consensus sequence\n");
369 fprintf(stderr, " -v print variants only (for -c)\n");
370 fprintf(stderr, " -g output in the GLFv3 format (suppressing -c/-i/-s)\n");
371 fprintf(stderr, " -T FLOAT theta in maq consensus calling model (for -c/-g) [%f]\n", d->c->theta);
372 fprintf(stderr, " -N INT number of haplotypes in the sample (for -c/-g) [%d]\n", d->c->n_hap);
373 fprintf(stderr, " -r FLOAT prior of a difference between two haplotypes (for -c/-g) [%f]\n", d->c->het_rate);
374 fprintf(stderr, " -G FLOAT prior of an indel between two haplotypes (for -c/-g) [%f]\n", d->ido->r_indel);
375 fprintf(stderr, " -I INT phred prob. of an indel in sequencing/prep. (for -c/-g) [%d]\n", d->ido->q_indel);
376 fprintf(stderr, "\n");
377 free(fn_list); free(fn_fa); free(d);
380 if (d->format & (BAM_PLF_RANBASE|BAM_PLF_1STBASE|BAM_PLF_ALLBASE)) d->format |= BAM_PLF_CNS;
381 if (fn_fa) d->fai = fai_load(fn_fa);
382 if (d->format & (BAM_PLF_CNS|BAM_PLF_GLF)) bam_maqcns_prepare(d->c); // consensus calling
383 if (d->format & BAM_PLF_GLF) { // for glf output
385 h = glf3_header_init();
386 d->fp_glf = bgzf_fdopen(fileno(stdout), "w");
387 glf3_header_write(d->fp_glf, h);
388 glf3_header_destroy(h);
390 if (d->fai == 0 && (d->format & (BAM_PLF_CNS|BAM_PLF_INDEL_ONLY)))
391 fprintf(stderr, "[bam_pileup] indels will not be called when -f is absent.\n");
392 if (fn_fa && is_SAM && fn_list == 0) fn_list = samfaipath(fn_fa);
396 fp = is_SAM? samopen(argv[optind], "r", fn_list) : samopen(argv[optind], "rb", 0);
397 if (fp == 0 || fp->header == 0) {
398 fprintf(stderr, "[bam_pileup] fail to read the header: non-exisiting file or wrong format.\n");
402 if (fn_pos) d->hash = load_pos(fn_pos, d->h);
403 sampileup(fp, d->mask, pileup_func, d);
404 samclose(fp); // d->h will be destroyed here
408 if (d->format & BAM_PLF_GLF) bgzf_close(d->fp_glf);
409 if (fn_pos) { // free the hash table
411 for (k = kh_begin(d->hash); k < kh_end(d->hash); ++k)
412 if (kh_exist(d->hash, k)) free(kh_val(d->hash, k));
413 kh_destroy(64, d->hash);
415 free(fn_pos); free(fn_list); free(fn_fa);
416 if (d->fai) fai_destroy(d->fai);
417 bam_maqcns_destroy(d->c);
418 free(d->ido); free(d->ref); free(d);