6 #include "bam_maqcns.h"
11 typedef int *indel_list_t;
12 KHASH_MAP_INIT_INT64(64, indel_list_t)
14 #define BAM_PLF_SIMPLE 0x01
15 #define BAM_PLF_CNS 0x02
16 #define BAM_PLF_INDEL_ONLY 0x04
17 #define BAM_PLF_GLF 0x08
22 bam_maqindel_opt_t *ido;
26 int tid, len, last_pos;
29 glfFile fp_glf; // for glf output only
32 char **__bam_get_lines(const char *fn, int *_n);
33 void bam_init_header_hash(bam_header_t *header);
34 int32_t bam_get_tid(const bam_header_t *header, const char *seq_name);
36 static khash_t(64) *load_pos(const char *fn, bam_header_t *h)
39 int i, j, n, *fields, max_fields;
41 bam_init_header_hash(h);
42 list = __bam_get_lines(fn, &n);
44 max_fields = 0; fields = 0;
45 for (i = 0; i < n; ++i) {
47 int chr, n_fields, ret;
50 n_fields = ksplit_core(str, 0, &max_fields, &fields);
51 if (n_fields < 2) continue;
52 chr = bam_get_tid(h, str + fields[0]);
54 fprintf(stderr, "[load_pos] unknown reference sequence name: %s\n", str + fields[0]);
57 x = (uint64_t)chr << 32 | (atoi(str + fields[1]) - 1);
58 k = kh_put(64, hash, x, &ret);
60 fprintf(stderr, "[load_pos] position %s:%s has been loaded.\n", str+fields[0], str+fields[1]);
66 for (j = 2; j < n_fields; ++j) {
67 char *s = str + fields[j];
68 if ((*s != '+' && *s != '-') || !isdigit(s[1])) break;
70 if (j > 2) { // update kh_val()
72 q = kh_val(hash, k) = (int*)calloc(j - 1, sizeof(int));
73 q[0] = j - 2; z = j; y = 1;
74 for (j = 2; j < z; ++j)
75 q[y++] = atoi(str + fields[j]);
80 free(list); free(fields);
84 // an analogy to pileup_func() below
85 static int glt3_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *pu, void *data)
87 pu_data_t *d = (pu_data_t*)data;
88 bam_maqindel_ret_t *r = 0;
89 int rb, *proposed_indels = 0;
94 fprintf(stderr, "[glt3_func] reference sequence is required for generating GLT. Abort!\n");
97 if (d->hash) { // only output a list of sites
98 khint_t k = kh_get(64, d->hash, (uint64_t)tid<<32|pos);
99 if (k == kh_end(d->hash)) return 0;
100 proposed_indels = kh_val(d->hash, k);
103 if (d->fai && (int)tid != d->tid) {
104 if (d->ref) { // then write the end mark
105 g3->rtype = GLF3_RTYPE_END;
106 glf3_write1(d->fp_glf, g3);
108 glf3_ref_write(d->fp_glf, d->h->target_name[tid], d->h->target_len[tid]); // write reference
110 d->ref = fai_fetch(d->fai, d->h->target_name[tid], &d->len);
114 rb = (d->ref && (int)pos < d->len)? d->ref[pos] : 'N';
115 g = bam_maqcns_glfgen(n, pu, bam_nt16_table[rb], d->c);
116 memcpy(g3, g, sizeof(glf1_t));
117 g3->rtype = GLF3_RTYPE_SUB;
118 g3->offset = pos - d->last_pos;
120 glf3_write1(d->fp_glf, g3);
122 r = bam_maqindel(n, pos, d->ido, pu, d->ref, proposed_indels[0], proposed_indels+1);
123 else r = bam_maqindel(n, pos, d->ido, pu, d->ref, 0, 0);
124 if (r) { // then write indel line
125 int het = 3 * n, min;
127 if (min > r->gl[0]) min = r->gl[0];
128 if (min > r->gl[1]) min = r->gl[1];
130 g3->rtype = GLF3_RTYPE_INDEL;
131 memset(g3->lk, 0, 10);
132 g3->lk[0] = r->gl[0] - min < 255? r->gl[0] - min : 255;
133 g3->lk[1] = r->gl[1] - min < 255? r->gl[1] - min : 255;
134 g3->lk[2] = het - min < 255? het - min : 255;
136 g3->indel_len[0] = r->indel1;
137 g3->indel_len[1] = r->indel2;
138 g3->min_lk = min < 255? min : 255;
139 g3->max_len = (abs(r->indel1) > abs(r->indel2)? abs(r->indel1) : abs(r->indel2)) + 1;
140 g3->indel_seq[0] = strdup(r->s[0]+1);
141 g3->indel_seq[1] = strdup(r->s[1]+1);
142 glf3_write1(d->fp_glf, g3);
143 bam_maqindel_ret_destroy(r);
150 static int pileup_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *pu, void *data)
152 pu_data_t *d = (pu_data_t*)data;
153 bam_maqindel_ret_t *r = 0;
154 int i, j, rb, rms_mapq = -1, max_mapq = 0, *proposed_indels = 0;
157 if (d->format & BAM_PLF_GLF) return glt3_func(tid, pos, n, pu, data);
158 if (d->hash) { // only output a list of sites
159 khint_t k = kh_get(64, d->hash, (uint64_t)tid<<32|pos);
160 if (k == kh_end(d->hash)) return 0;
161 proposed_indels = kh_val(d->hash, k);
163 if (d->fai && (int)tid != d->tid) { // then update d->ref
165 d->ref = fai_fetch(d->fai, d->h->target_name[tid], &d->len);
168 rb = (d->ref && (int)pos < d->len)? d->ref[pos] : 'N';
169 if (d->format & BAM_PLF_INDEL_ONLY) {
170 for (i = 0; i < n; ++i)
171 if (pu[i].indel != 0) break;
172 if (i == n) return 0;
174 printf("%s\t%d\t%c\t", d->h->target_name[tid], pos + 1, rb);
175 if (d->format & BAM_PLF_CNS) { // consensus
176 int ref_q, rb4 = bam_nt16_table[rb];
177 x = bam_maqcns_call(n, pu, d->c);
179 if (rb4 != 15 && x>>28 != 15 && x>>28 != rb4) { // a SNP
180 ref_q = ((x>>24&0xf) == rb4)? x>>8&0xff : (x>>8&0xff) + (x&0xff);
181 if (ref_q > 255) ref_q = 255;
183 printf("%c\t%d\t%d\t%d\t", bam_nt16_rev_table[x>>28], x>>8&0xff, ref_q, x>>16&0xff);
184 rms_mapq = x>>16&0xff;
186 if ((d->format & (BAM_PLF_CNS|BAM_PLF_INDEL_ONLY)) && d->ref) {
188 r = bam_maqindel(n, pos, d->ido, pu, d->ref, proposed_indels[0], proposed_indels+1);
189 else r = bam_maqindel(n, pos, d->ido, pu, d->ref, 0, 0);
193 for (i = 0; i < n; ++i) {
194 const bam_pileup1_t *p = pu + i;
195 if (max_mapq < p->b->core.qual) max_mapq = p->b->core.qual;
196 if (p->is_head) printf("^%c", p->b->core.qual > 93? 126 : p->b->core.qual + 33);
198 int c = bam_nt16_rev_table[bam1_seqi(bam1_seq(p->b), p->qpos)];
199 if (c == '=' || toupper(c) == toupper(rb)) c = bam1_strand(p->b)? ',' : '.';
200 else c = bam1_strand(p->b)? tolower(c) : toupper(c);
203 printf("+%d", p->indel);
204 for (j = 1; j <= p->indel; ++j) {
205 c = bam_nt16_rev_table[bam1_seqi(bam1_seq(p->b), p->qpos + j)];
206 putchar(bam1_strand(p->b)? tolower(c) : toupper(c));
208 } else if (p->indel < 0) {
209 printf("%d", p->indel);
210 for (j = 1; j <= -p->indel; ++j) {
211 c = (d->ref && (int)pos+j < d->len)? d->ref[pos+j] : 'N';
212 putchar(bam1_strand(p->b)? tolower(c) : toupper(c));
216 if (p->is_tail) putchar('$');
219 for (i = 0; i < n; ++i) {
220 const bam_pileup1_t *p = pu + i;
221 int c = bam1_qual(p->b)[p->qpos] + 33;
222 if (c > 126) c = 126;
225 if (d->format & BAM_PLF_SIMPLE) {
227 for (i = 0; i < n; ++i) {
228 int c = pu[i].b->core.qual + 33;
229 if (c > 126) c = 126;
234 if (r) { // then print indel line
235 printf("%s\t%d\t*\t", d->h->target_name[tid], pos + 1);
236 if (r->gt < 2) printf("%s/%s\t", r->s[r->gt], r->s[r->gt]);
237 else printf("%s/%s\t", r->s[0], r->s[1]);
238 printf("%d\t%d\t", r->q_cns, r->q_ref);
239 printf("%d\t%d\t", rms_mapq >= 0? rms_mapq : max_mapq, n);
240 printf("%s\t%s\t", r->s[0], r->s[1]);
241 //printf("%d\t%d\t", r->gl[0], r->gl[1]);
242 printf("%d\t%d\t%d\n", r->cnt1, r->cnt2, r->cnt_anti);
243 bam_maqindel_ret_destroy(r);
248 int bam_pileup(int argc, char *argv[])
251 char *fn_list = 0, *fn_fa = 0, *fn_pos = 0;
252 pu_data_t *d = (pu_data_t*)calloc(1, sizeof(pu_data_t));
253 d->tid = -1; d->mask = BAM_DEF_MASK;
254 d->c = bam_maqcns_init();
255 d->ido = bam_maqindel_opt_init();
256 while ((c = getopt(argc, argv, "st:f:cT:N:r:l:im:gI:G:")) >= 0) {
258 case 's': d->format |= BAM_PLF_SIMPLE; break;
259 case 't': fn_list = strdup(optarg); break;
260 case 'l': fn_pos = strdup(optarg); break;
261 case 'f': fn_fa = strdup(optarg); break;
262 case 'T': d->c->theta = atof(optarg); break;
263 case 'N': d->c->n_hap = atoi(optarg); break;
264 case 'r': d->c->het_rate = atof(optarg); break;
265 case 'c': d->format |= BAM_PLF_CNS; break;
266 case 'i': d->format |= BAM_PLF_INDEL_ONLY; break;
267 case 'm': d->mask = strtol(optarg, 0, 0); break;
268 case 'g': d->format |= BAM_PLF_GLF; break;
269 case 'I': d->ido->q_indel = atoi(optarg); break;
270 case 'G': d->ido->r_indel = atof(optarg); break;
271 default: fprintf(stderr, "Unrecognizd option '-%c'.\n", c); return 1;
274 if (optind == argc) {
275 fprintf(stderr, "\n");
276 fprintf(stderr, "Usage: samtools pileup [options] <in.bam>|<in.sam>\n\n");
277 fprintf(stderr, "Option: -s simple (yet incomplete) pileup format\n");
278 fprintf(stderr, " -i only show lines/consensus with indels\n");
279 fprintf(stderr, " -m INT filtering reads with bits in INT [%d]\n", d->mask);
280 fprintf(stderr, " -t FILE list of reference sequences (assume the input is in SAM)\n");
281 fprintf(stderr, " -l FILE list of sites at which pileup is output\n");
282 fprintf(stderr, " -f FILE reference sequence in the FASTA format\n\n");
283 fprintf(stderr, " -c output the maq consensus sequence\n");
284 fprintf(stderr, " -g output in the GLFv3 format (suppressing -c/-i/-s)\n");
285 fprintf(stderr, " -T FLOAT theta in maq consensus calling model (for -c/-g) [%f]\n", d->c->theta);
286 fprintf(stderr, " -N INT number of haplotypes in the sample (for -c/-g) [%d]\n", d->c->n_hap);
287 fprintf(stderr, " -r FLOAT prior of a difference between two haplotypes (for -c/-g) [%f]\n", d->c->het_rate);
288 fprintf(stderr, " -G FLOAT prior of an indel between two haplotypes (for -c/-g) [%f]\n", d->ido->r_indel);
289 fprintf(stderr, " -I INT phred prob. of an indel in sequencing/prep. (for -c/-g) [%d]\n", d->ido->q_indel);
290 fprintf(stderr, "\n");
291 free(fn_list); free(fn_fa); free(d);
294 if (fn_fa) d->fai = fai_load(fn_fa);
295 if (d->format & (BAM_PLF_CNS|BAM_PLF_GLF)) bam_maqcns_prepare(d->c); // consensus calling
296 if (d->format & BAM_PLF_GLF) { // for glf output
298 h = glf3_header_init();
299 d->fp_glf = bgzf_fdopen(fileno(stdout), "w");
300 glf3_header_write(d->fp_glf, h);
301 glf3_header_destroy(h);
303 if (d->fai == 0 && (d->format & (BAM_PLF_CNS|BAM_PLF_INDEL_ONLY)))
304 fprintf(stderr, "[bam_pileup] indels will not be called when -f is absent.\n");
307 fp = fn_list? samopen(argv[optind], "r", fn_list) : samopen(argv[optind], "rb", 0);
308 if (fp == 0 || fp->header == 0) {
309 fprintf(stderr, "[bam_pileup] fail to read the header: non-exisiting file or wrong format.\n");
313 if (fn_pos) d->hash = load_pos(fn_pos, d->h);
314 sampileup(fp, d->mask, pileup_func, d);
315 samclose(fp); // d->h will be destroyed here
319 if (d->format & BAM_PLF_GLF) bgzf_close(d->fp_glf);
320 if (fn_pos) { // free the hash table
322 for (k = kh_begin(d->hash); k < kh_end(d->hash); ++k)
323 if (kh_exist(d->hash, k)) free(kh_val(d->hash, k));
324 kh_destroy(64, d->hash);
326 free(fn_pos); free(fn_list); free(fn_fa);
327 if (d->fai) fai_destroy(d->fai);
328 bam_maqcns_destroy(d->c);
329 free(d->ido); free(d->ref); free(d);