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
24 bam_maqindel_opt_t *ido;
28 int tid, len, last_pos;
31 glfFile fp_glf; // for glf output only
34 char **__bam_get_lines(const char *fn, int *_n);
35 void bam_init_header_hash(bam_header_t *header);
36 int32_t bam_get_tid(const bam_header_t *header, const char *seq_name);
38 static khash_t(64) *load_pos(const char *fn, bam_header_t *h)
41 int i, j, n, *fields, max_fields;
43 bam_init_header_hash(h);
44 list = __bam_get_lines(fn, &n);
46 max_fields = 0; fields = 0;
47 for (i = 0; i < n; ++i) {
49 int chr, n_fields, ret;
52 n_fields = ksplit_core(str, 0, &max_fields, &fields);
53 if (n_fields < 2) continue;
54 chr = bam_get_tid(h, str + fields[0]);
56 fprintf(stderr, "[load_pos] unknown reference sequence name: %s\n", str + fields[0]);
59 x = (uint64_t)chr << 32 | (atoi(str + fields[1]) - 1);
60 k = kh_put(64, hash, x, &ret);
62 fprintf(stderr, "[load_pos] position %s:%s has been loaded.\n", str+fields[0], str+fields[1]);
68 for (j = 2; j < n_fields; ++j) {
69 char *s = str + fields[j];
70 if ((*s != '+' && *s != '-') || !isdigit(s[1])) break;
72 if (j > 2) { // update kh_val()
74 q = kh_val(hash, k) = (int*)calloc(j - 1, sizeof(int));
75 q[0] = j - 2; z = j; y = 1;
76 for (j = 2; j < z; ++j)
77 q[y++] = atoi(str + fields[j]);
82 free(list); free(fields);
86 // an analogy to pileup_func() below
87 static int glt3_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *pu, void *data)
89 pu_data_t *d = (pu_data_t*)data;
90 bam_maqindel_ret_t *r = 0;
91 int rb, *proposed_indels = 0;
96 fprintf(stderr, "[glt3_func] reference sequence is required for generating GLT. Abort!\n");
99 if (d->hash) { // only output a list of sites
100 khint_t k = kh_get(64, d->hash, (uint64_t)tid<<32|pos);
101 if (k == kh_end(d->hash)) return 0;
102 proposed_indels = kh_val(d->hash, k);
105 if (d->fai && (int)tid != d->tid) {
106 if (d->ref) { // then write the end mark
107 g3->rtype = GLF3_RTYPE_END;
108 glf3_write1(d->fp_glf, g3);
110 glf3_ref_write(d->fp_glf, d->h->target_name[tid], d->h->target_len[tid]); // write reference
112 d->ref = fai_fetch(d->fai, d->h->target_name[tid], &d->len);
116 rb = (d->ref && (int)pos < d->len)? d->ref[pos] : 'N';
117 g = bam_maqcns_glfgen(n, pu, bam_nt16_table[rb], d->c);
118 memcpy(g3, g, sizeof(glf1_t));
119 g3->rtype = GLF3_RTYPE_SUB;
120 g3->offset = pos - d->last_pos;
122 glf3_write1(d->fp_glf, g3);
124 r = bam_maqindel(n, pos, d->ido, pu, d->ref, proposed_indels[0], proposed_indels+1);
125 else r = bam_maqindel(n, pos, d->ido, pu, d->ref, 0, 0);
126 if (r) { // then write indel line
127 int het = 3 * n, min;
129 if (min > r->gl[0]) min = r->gl[0];
130 if (min > r->gl[1]) min = r->gl[1];
132 g3->rtype = GLF3_RTYPE_INDEL;
133 memset(g3->lk, 0, 10);
134 g3->lk[0] = r->gl[0] - min < 255? r->gl[0] - min : 255;
135 g3->lk[1] = r->gl[1] - min < 255? r->gl[1] - min : 255;
136 g3->lk[2] = het - min < 255? het - min : 255;
138 g3->indel_len[0] = r->indel1;
139 g3->indel_len[1] = r->indel2;
140 g3->min_lk = min < 255? min : 255;
141 g3->max_len = (abs(r->indel1) > abs(r->indel2)? abs(r->indel1) : abs(r->indel2)) + 1;
142 g3->indel_seq[0] = strdup(r->s[0]+1);
143 g3->indel_seq[1] = strdup(r->s[1]+1);
144 glf3_write1(d->fp_glf, g3);
145 bam_maqindel_ret_destroy(r);
152 static int pileup_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *pu, void *data)
154 pu_data_t *d = (pu_data_t*)data;
155 bam_maqindel_ret_t *r = 0;
156 int i, j, rb, rms_mapq = -1, *proposed_indels = 0;
160 // if GLF is required, suppress -c completely
161 if (d->format & BAM_PLF_GLF) return glt3_func(tid, pos, n, pu, data);
162 // if d->hash is initialized, only output the sites in the hash table
164 khint_t k = kh_get(64, d->hash, (uint64_t)tid<<32|pos);
165 if (k == kh_end(d->hash)) return 0;
166 proposed_indels = kh_val(d->hash, k);
168 // update d->ref if necessary
169 if (d->fai && (int)tid != d->tid) {
171 d->ref = fai_fetch(d->fai, d->h->target_name[tid], &d->len);
174 rb = (d->ref && (int)pos < d->len)? d->ref[pos] : 'N';
175 // when the indel-only mode is asked for, return if no reads mapped with indels
176 if (d->format & BAM_PLF_INDEL_ONLY) {
177 for (i = 0; i < n; ++i)
178 if (pu[i].indel != 0) break;
179 if (i == n) return 0;
181 // call the consensus and indel
182 if (d->format & BAM_PLF_CNS) // call consensus
183 cns = bam_maqcns_call(n, pu, d->c);
184 if ((d->format & (BAM_PLF_CNS|BAM_PLF_INDEL_ONLY)) && d->ref) { // call indels
185 if (proposed_indels) // the first element gives the size of the array
186 r = bam_maqindel(n, pos, d->ido, pu, d->ref, proposed_indels[0], proposed_indels+1);
187 else r = bam_maqindel(n, pos, d->ido, pu, d->ref, 0, 0);
189 // when only variant sites are asked for, test if the site is a variant
190 if ((d->format & BAM_PLF_CNS) && (d->format & BAM_PLF_VAR_ONLY)) {
191 if (!(bam_nt16_table[rb] != 15 && cns>>28 != bam_nt16_table[rb])) { // not a SNP
192 if (!(r && (r->gt == 2 || strcmp(r->s[r->gt], "*")))) { // not an indel
193 if (r) bam_maqindel_ret_destroy(r);
198 // print the first 3 columns
199 printf("%s\t%d\t%c\t", d->h->target_name[tid], pos + 1, rb);
200 // print consensus information if required
201 if (d->format & BAM_PLF_CNS) {
202 int ref_q, rb4 = bam_nt16_table[rb];
204 if (rb4 != 15 && cns>>28 != 15 && cns>>28 != rb4) { // a SNP
205 ref_q = ((cns>>24&0xf) == rb4)? cns>>8&0xff : (cns>>8&0xff) + (cns&0xff);
206 if (ref_q > 255) ref_q = 255;
208 rms_mapq = cns>>16&0xff;
209 printf("%c\t%d\t%d\t%d\t", bam_nt16_rev_table[cns>>28], cns>>8&0xff, ref_q, rms_mapq);
211 // print pileup sequences
213 rms_aux = 0; // we need to recalculate rms_mapq when -c is not flagged on the command line
214 for (i = 0; i < n; ++i) {
215 const bam_pileup1_t *p = pu + i;
216 int tmp = p->b->core.qual < d->c->cap_mapQ? p->b->core.qual : d->c->cap_mapQ;
217 rms_aux += tmp * tmp;
218 if (p->is_head) printf("^%c", p->b->core.qual > 93? 126 : p->b->core.qual + 33);
220 int c = bam_nt16_rev_table[bam1_seqi(bam1_seq(p->b), p->qpos)];
221 if (c == '=' || toupper(c) == toupper(rb)) c = bam1_strand(p->b)? ',' : '.';
222 else c = bam1_strand(p->b)? tolower(c) : toupper(c);
225 printf("+%d", p->indel);
226 for (j = 1; j <= p->indel; ++j) {
227 c = bam_nt16_rev_table[bam1_seqi(bam1_seq(p->b), p->qpos + j)];
228 putchar(bam1_strand(p->b)? tolower(c) : toupper(c));
230 } else if (p->indel < 0) {
231 printf("%d", p->indel);
232 for (j = 1; j <= -p->indel; ++j) {
233 c = (d->ref && (int)pos+j < d->len)? d->ref[pos+j] : 'N';
234 putchar(bam1_strand(p->b)? tolower(c) : toupper(c));
238 if (p->is_tail) putchar('$');
241 rms_aux = (uint64_t)(sqrt((double)rms_aux / n) + .499);
242 if (rms_mapq < 0) rms_mapq = rms_aux;
245 for (i = 0; i < n; ++i) {
246 const bam_pileup1_t *p = pu + i;
247 int c = bam1_qual(p->b)[p->qpos] + 33;
248 if (c > 126) c = 126;
251 // print mapping quality if -s is flagged on the command line
252 if (d->format & BAM_PLF_SIMPLE) {
254 for (i = 0; i < n; ++i) {
255 int c = pu[i].b->core.qual + 33;
256 if (c > 126) c = 126;
261 // print the indel line if r has been calculated. This only happens if:
262 // a) -c or -i are flagged, AND b) the reference sequence is available
264 printf("%s\t%d\t*\t", d->h->target_name[tid], pos + 1);
265 if (r->gt < 2) printf("%s/%s\t", r->s[r->gt], r->s[r->gt]);
266 else printf("%s/%s\t", r->s[0], r->s[1]);
267 printf("%d\t%d\t", r->q_cns, r->q_ref);
268 printf("%d\t%d\t", rms_mapq, n);
269 printf("%s\t%s\t", r->s[0], r->s[1]);
270 //printf("%d\t%d\t", r->gl[0], r->gl[1]);
271 printf("%d\t%d\t%d\n", r->cnt1, r->cnt2, r->cnt_anti);
272 bam_maqindel_ret_destroy(r);
277 int bam_pileup(int argc, char *argv[])
280 char *fn_list = 0, *fn_fa = 0, *fn_pos = 0;
281 pu_data_t *d = (pu_data_t*)calloc(1, sizeof(pu_data_t));
282 d->tid = -1; d->mask = BAM_DEF_MASK;
283 d->c = bam_maqcns_init();
284 d->ido = bam_maqindel_opt_init();
285 while ((c = getopt(argc, argv, "st:f:cT:N:r:l:im:gI:G:vM:S")) >= 0) {
287 case 's': d->format |= BAM_PLF_SIMPLE; break;
288 case 't': fn_list = strdup(optarg); break;
289 case 'l': fn_pos = strdup(optarg); break;
290 case 'f': fn_fa = strdup(optarg); break;
291 case 'T': d->c->theta = atof(optarg); break;
292 case 'N': d->c->n_hap = atoi(optarg); break;
293 case 'r': d->c->het_rate = atof(optarg); break;
294 case 'M': d->c->cap_mapQ = atoi(optarg); break;
295 case 'c': d->format |= BAM_PLF_CNS; break;
296 case 'i': d->format |= BAM_PLF_INDEL_ONLY; break;
297 case 'v': d->format |= BAM_PLF_VAR_ONLY; break;
298 case 'm': d->mask = strtol(optarg, 0, 0); break;
299 case 'g': d->format |= BAM_PLF_GLF; break;
300 case 'I': d->ido->q_indel = atoi(optarg); break;
301 case 'G': d->ido->r_indel = atof(optarg); break;
302 case 'S': is_SAM = 1; break;
303 default: fprintf(stderr, "Unrecognizd option '-%c'.\n", c); return 1;
306 if (fn_list) is_SAM = 1;
307 if (optind == argc) {
308 fprintf(stderr, "\n");
309 fprintf(stderr, "Usage: samtools pileup [options] <in.bam>|<in.sam>\n\n");
310 fprintf(stderr, "Option: -s simple (yet incomplete) pileup format\n");
311 fprintf(stderr, " -S the input is in SAM\n");
312 fprintf(stderr, " -i only show lines/consensus with indels\n");
313 fprintf(stderr, " -m INT filtering reads with bits in INT [%d]\n", d->mask);
314 fprintf(stderr, " -M INT cap mapping quality at INT [%d]\n", d->c->cap_mapQ);
315 fprintf(stderr, " -t FILE list of reference sequences (assume the input is in SAM)\n");
316 fprintf(stderr, " -l FILE list of sites at which pileup is output\n");
317 fprintf(stderr, " -f FILE reference sequence in the FASTA format\n\n");
318 fprintf(stderr, " -c output the maq consensus sequence\n");
319 fprintf(stderr, " -v print variants only (for -c)\n");
320 fprintf(stderr, " -g output in the GLFv3 format (suppressing -c/-i/-s)\n");
321 fprintf(stderr, " -T FLOAT theta in maq consensus calling model (for -c/-g) [%f]\n", d->c->theta);
322 fprintf(stderr, " -N INT number of haplotypes in the sample (for -c/-g) [%d]\n", d->c->n_hap);
323 fprintf(stderr, " -r FLOAT prior of a difference between two haplotypes (for -c/-g) [%f]\n", d->c->het_rate);
324 fprintf(stderr, " -G FLOAT prior of an indel between two haplotypes (for -c/-g) [%f]\n", d->ido->r_indel);
325 fprintf(stderr, " -I INT phred prob. of an indel in sequencing/prep. (for -c/-g) [%d]\n", d->ido->q_indel);
326 fprintf(stderr, "\n");
327 free(fn_list); free(fn_fa); free(d);
330 if (fn_fa) d->fai = fai_load(fn_fa);
331 if (d->format & (BAM_PLF_CNS|BAM_PLF_GLF)) bam_maqcns_prepare(d->c); // consensus calling
332 if (d->format & BAM_PLF_GLF) { // for glf output
334 h = glf3_header_init();
335 d->fp_glf = bgzf_fdopen(fileno(stdout), "w");
336 glf3_header_write(d->fp_glf, h);
337 glf3_header_destroy(h);
339 if (d->fai == 0 && (d->format & (BAM_PLF_CNS|BAM_PLF_INDEL_ONLY)))
340 fprintf(stderr, "[bam_pileup] indels will not be called when -f is absent.\n");
343 fp = is_SAM? samopen(argv[optind], "r", fn_list) : samopen(argv[optind], "rb", 0);
344 if (fp == 0 || fp->header == 0) {
345 fprintf(stderr, "[bam_pileup] fail to read the header: non-exisiting file or wrong format.\n");
349 if (fn_pos) d->hash = load_pos(fn_pos, d->h);
350 sampileup(fp, d->mask, pileup_func, d);
351 samclose(fp); // d->h will be destroyed here
355 if (d->format & BAM_PLF_GLF) bgzf_close(d->fp_glf);
356 if (fn_pos) { // free the hash table
358 for (k = kh_begin(d->hash); k < kh_end(d->hash); ++k)
359 if (kh_exist(d->hash, k)) free(kh_val(d->hash, k));
360 kh_destroy(64, d->hash);
362 free(fn_pos); free(fn_list); free(fn_fa);
363 if (d->fai) fai_destroy(d->fai);
364 bam_maqcns_destroy(d->c);
365 free(d->ido); free(d->ref); free(d);