Mercurial > repos > lsong10 > psiclass

comparison PsiCLASS-1.0.2/samtools-0.1.19/bam_md.c @ 0:903fc43d6227 draft default tip

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Uploaded
author lsong10
date Fri, 26 Mar 2021 16:52:45 +0000
parents
children
comparison
equal deleted inserted replaced
-1:000000000000 0:903fc43d6227
1 #include <unistd.h>
2 #include <assert.h>
3 #include <string.h>
4 #include <ctype.h>
5 #include <math.h>
6 #include "faidx.h"
7 #include "sam.h"
8 #include "kstring.h"
9 #include "kaln.h"
10 #include "kprobaln.h"
11
12 #define USE_EQUAL 1
13 #define DROP_TAG 2
14 #define BIN_QUAL 4
15 #define UPDATE_NM 8
16 #define UPDATE_MD 16
17 #define HASH_QNM 32
18
19 char bam_nt16_nt4_table[] = { 4, 0, 1, 4, 2, 4, 4, 4, 3, 4, 4, 4, 4, 4, 4, 4 };
20
21 int bam_aux_drop_other(bam1_t *b, uint8_t *s);
22
23 void bam_fillmd1_core(bam1_t *b, char *ref, int flag, int max_nm)
24 {
25 uint8_t *seq = bam1_seq(b);
26 uint32_t *cigar = bam1_cigar(b);
27 bam1_core_t *c = &b->core;
28 int i, x, y, u = 0;
29 kstring_t *str;
30 int32_t old_nm_i = -1, nm = 0;
31
32 str = (kstring_t*)calloc(1, sizeof(kstring_t));
33 for (i = y = 0, x = c->pos; i < c->n_cigar; ++i) {
34 int j, l = cigar[i]>>4, op = cigar[i]&0xf;
35 if (op == BAM_CMATCH || op == BAM_CEQUAL || op == BAM_CDIFF) {
36 for (j = 0; j < l; ++j) {
37 int z = y + j;
38 int c1 = bam1_seqi(seq, z), c2 = bam_nt16_table[(int)ref[x+j]];
39 if (ref[x+j] == 0) break; // out of boundary
40 if ((c1 == c2 && c1 != 15 && c2 != 15) || c1 == 0) { // a match
41 if (flag&USE_EQUAL) seq[z/2] &= (z&1)? 0xf0 : 0x0f;
42 ++u;
43 } else {
44 kputw(u, str); kputc(ref[x+j], str);
45 u = 0; ++nm;
46 }
47 }
48 if (j < l) break;
49 x += l; y += l;
50 } else if (op == BAM_CDEL) {
51 kputw(u, str); kputc('^', str);
52 for (j = 0; j < l; ++j) {
53 if (ref[x+j] == 0) break;
54 kputc(ref[x+j], str);
55 }
56 u = 0;
57 if (j < l) break;
58 x += l; nm += l;
59 } else if (op == BAM_CINS || op == BAM_CSOFT_CLIP) {
60 y += l;
61 if (op == BAM_CINS) nm += l;
62 } else if (op == BAM_CREF_SKIP) {
63 x += l;
64 }
65 }
66 kputw(u, str);
67 // apply max_nm
68 if (max_nm > 0 && nm >= max_nm) {
69 for (i = y = 0, x = c->pos; i < c->n_cigar; ++i) {
70 int j, l = cigar[i]>>4, op = cigar[i]&0xf;
71 if (op == BAM_CMATCH || op == BAM_CEQUAL || op == BAM_CDIFF) {
72 for (j = 0; j < l; ++j) {
73 int z = y + j;
74 int c1 = bam1_seqi(seq, z), c2 = bam_nt16_table[(int)ref[x+j]];
75 if (ref[x+j] == 0) break; // out of boundary
76 if ((c1 == c2 && c1 != 15 && c2 != 15) || c1 == 0) { // a match
77 seq[z/2] |= (z&1)? 0x0f : 0xf0;
78 bam1_qual(b)[z] = 0;
79 }
80 }
81 if (j < l) break;
82 x += l; y += l;
83 } else if (op == BAM_CDEL || op == BAM_CREF_SKIP) x += l;
84 else if (op == BAM_CINS || op == BAM_CSOFT_CLIP) y += l;
85 }
86 }
87 // update NM
88 if (flag & UPDATE_NM) {
89 uint8_t *old_nm = bam_aux_get(b, "NM");
90 if (c->flag & BAM_FUNMAP) return;
91 if (old_nm) old_nm_i = bam_aux2i(old_nm);
92 if (!old_nm) bam_aux_append(b, "NM", 'i', 4, (uint8_t*)&nm);
93 else if (nm != old_nm_i) {
94 fprintf(stderr, "[bam_fillmd1] different NM for read '%s': %d -> %d\n", bam1_qname(b), old_nm_i, nm);
95 bam_aux_del(b, old_nm);
96 bam_aux_append(b, "NM", 'i', 4, (uint8_t*)&nm);
97 }
98 }
99 // update MD
100 if (flag & UPDATE_MD) {
101 uint8_t *old_md = bam_aux_get(b, "MD");
102 if (c->flag & BAM_FUNMAP) return;
103 if (!old_md) bam_aux_append(b, "MD", 'Z', str->l + 1, (uint8_t*)str->s);
104 else {
105 int is_diff = 0;
106 if (strlen((char*)old_md+1) == str->l) {
107 for (i = 0; i < str->l; ++i)
108 if (toupper(old_md[i+1]) != toupper(str->s[i]))
109 break;
110 if (i < str->l) is_diff = 1;
111 } else is_diff = 1;
112 if (is_diff) {
113 fprintf(stderr, "[bam_fillmd1] different MD for read '%s': '%s' -> '%s'\n", bam1_qname(b), old_md+1, str->s);
114 bam_aux_del(b, old_md);
115 bam_aux_append(b, "MD", 'Z', str->l + 1, (uint8_t*)str->s);
116 }
117 }
118 }
119 // drop all tags but RG
120 if (flag&DROP_TAG) {
121 uint8_t *q = bam_aux_get(b, "RG");
122 bam_aux_drop_other(b, q);
123 }
124 // reduce the resolution of base quality
125 if (flag&BIN_QUAL) {
126 uint8_t *qual = bam1_qual(b);
127 for (i = 0; i < b->core.l_qseq; ++i)
128 if (qual[i] >= 3) qual[i] = qual[i]/10*10 + 7;
129 }
130 free(str->s); free(str);
131 }
132
133 void bam_fillmd1(bam1_t *b, char *ref, int flag)
134 {
135 bam_fillmd1_core(b, ref, flag, 0);
136 }
137
138 int bam_cap_mapQ(bam1_t *b, char *ref, int thres)
139 {
140 uint8_t *seq = bam1_seq(b), *qual = bam1_qual(b);
141 uint32_t *cigar = bam1_cigar(b);
142 bam1_core_t *c = &b->core;
143 int i, x, y, mm, q, len, clip_l, clip_q;
144 double t;
145 if (thres < 0) thres = 40; // set the default
146 mm = q = len = clip_l = clip_q = 0;
147 for (i = y = 0, x = c->pos; i < c->n_cigar; ++i) {
148 int j, l = cigar[i]>>4, op = cigar[i]&0xf;
149 if (op == BAM_CMATCH || op == BAM_CEQUAL || op == BAM_CDIFF) {
150 for (j = 0; j < l; ++j) {
151 int z = y + j;
152 int c1 = bam1_seqi(seq, z), c2 = bam_nt16_table[(int)ref[x+j]];
153 if (ref[x+j] == 0) break; // out of boundary
154 if (c2 != 15 && c1 != 15 && qual[z] >= 13) { // not ambiguous
155 ++len;
156 if (c1 && c1 != c2 && qual[z] >= 13) { // mismatch
157 ++mm;
158 q += qual[z] > 33? 33 : qual[z];
159 }
160 }
161 }
162 if (j < l) break;
163 x += l; y += l; len += l;
164 } else if (op == BAM_CDEL) {
165 for (j = 0; j < l; ++j)
166 if (ref[x+j] == 0) break;
167 if (j < l) break;
168 x += l;
169 } else if (op == BAM_CSOFT_CLIP) {
170 for (j = 0; j < l; ++j) clip_q += qual[y+j];
171 clip_l += l;
172 y += l;
173 } else if (op == BAM_CHARD_CLIP) {
174 clip_q += 13 * l;
175 clip_l += l;
176 } else if (op == BAM_CINS) y += l;
177 else if (op == BAM_CREF_SKIP) x += l;
178 }
179 for (i = 0, t = 1; i < mm; ++i)
180 t *= (double)len / (i+1);
181 t = q - 4.343 * log(t) + clip_q / 5.;
182 if (t > thres) return -1;
183 if (t < 0) t = 0;
184 t = sqrt((thres - t) / thres) * thres;
185 // fprintf(stderr, "%s %lf %d\n", bam1_qname(b), t, q);
186 return (int)(t + .499);
187 }
188
189 int bam_prob_realn_core(bam1_t *b, const char *ref, int flag)
190 {
191 int k, i, bw, x, y, yb, ye, xb, xe, apply_baq = flag&1, extend_baq = flag>>1&1, redo_baq = flag&4;
192 uint32_t *cigar = bam1_cigar(b);
193 bam1_core_t *c = &b->core;
194 kpa_par_t conf = kpa_par_def;
195 uint8_t *bq = 0, *zq = 0, *qual = bam1_qual(b);
196 if ((c->flag & BAM_FUNMAP) || b->core.l_qseq == 0) return -1; // do nothing
197 // test if BQ or ZQ is present
198 if ((bq = bam_aux_get(b, "BQ")) != 0) ++bq;
199 if ((zq = bam_aux_get(b, "ZQ")) != 0 && *zq == 'Z') ++zq;
200 if (bq && redo_baq)
201 {
202 bam_aux_del(b, bq-1);
203 bq = 0;
204 }
205 if (bq && zq) { // remove the ZQ tag
206 bam_aux_del(b, zq-1);
207 zq = 0;
208 }
209 if (bq || zq) {
210 if ((apply_baq && zq) || (!apply_baq && bq)) return -3; // in both cases, do nothing
211 if (bq && apply_baq) { // then convert BQ to ZQ
212 for (i = 0; i < c->l_qseq; ++i)
213 qual[i] = qual[i] + 64 < bq[i]? 0 : qual[i] - ((int)bq[i] - 64);
214 *(bq - 3) = 'Z';
215 } else if (zq && !apply_baq) { // then convert ZQ to BQ
216 for (i = 0; i < c->l_qseq; ++i)
217 qual[i] += (int)zq[i] - 64;
218 *(zq - 3) = 'B';
219 }
220 return 0;
221 }
222 // find the start and end of the alignment
223 x = c->pos, y = 0, yb = ye = xb = xe = -1;
224 for (k = 0; k < c->n_cigar; ++k) {
225 int op, l;
226 op = cigar[k]&0xf; l = cigar[k]>>4;
227 if (op == BAM_CMATCH || op == BAM_CEQUAL || op == BAM_CDIFF) {
228 if (yb < 0) yb = y;
229 if (xb < 0) xb = x;
230 ye = y + l; xe = x + l;
231 x += l; y += l;
232 } else if (op == BAM_CSOFT_CLIP || op == BAM_CINS) y += l;
233 else if (op == BAM_CDEL) x += l;
234 else if (op == BAM_CREF_SKIP) return -1; // do nothing if there is a reference skip
235 }
236 // set bandwidth and the start and the end
237 bw = 7;
238 if (abs((xe - xb) - (ye - yb)) > bw)
239 bw = abs((xe - xb) - (ye - yb)) + 3;
240 conf.bw = bw;
241 xb -= yb + bw/2; if (xb < 0) xb = 0;
242 xe += c->l_qseq - ye + bw/2;
243 if (xe - xb - c->l_qseq > bw)
244 xb += (xe - xb - c->l_qseq - bw) / 2, xe -= (xe - xb - c->l_qseq - bw) / 2;
245 { // glocal
246 uint8_t *s, *r, *q, *seq = bam1_seq(b), *bq;
247 int *state;
248 bq = calloc(c->l_qseq + 1, 1);
249 memcpy(bq, qual, c->l_qseq);
250 s = calloc(c->l_qseq, 1);
251 for (i = 0; i < c->l_qseq; ++i) s[i] = bam_nt16_nt4_table[bam1_seqi(seq, i)];
252 r = calloc(xe - xb, 1);
253 for (i = xb; i < xe; ++i) {
254 if (ref[i] == 0) { xe = i; break; }
255 r[i-xb] = bam_nt16_nt4_table[bam_nt16_table[(int)ref[i]]];
256 }
257 state = calloc(c->l_qseq, sizeof(int));
258 q = calloc(c->l_qseq, 1);
259 kpa_glocal(r, xe-xb, s, c->l_qseq, qual, &conf, state, q);
260 if (!extend_baq) { // in this block, bq[] is capped by base quality qual[]
261 for (k = 0, x = c->pos, y = 0; k < c->n_cigar; ++k) {
262 int op = cigar[k]&0xf, l = cigar[k]>>4;
263 if (op == BAM_CMATCH || op == BAM_CEQUAL || op == BAM_CDIFF) {
264 for (i = y; i < y + l; ++i) {
265 if ((state[i]&3) != 0 || state[i]>>2 != x - xb + (i - y)) bq[i] = 0;
266 else bq[i] = bq[i] < q[i]? bq[i] : q[i];
267 }
268 x += l; y += l;
269 } else if (op == BAM_CSOFT_CLIP || op == BAM_CINS) y += l;
270 else if (op == BAM_CDEL) x += l;
271 }
272 for (i = 0; i < c->l_qseq; ++i) bq[i] = qual[i] - bq[i] + 64; // finalize BQ
273 } else { // in this block, bq[] is BAQ that can be larger than qual[] (different from the above!)
274 uint8_t *left, *rght;
275 left = calloc(c->l_qseq, 1); rght = calloc(c->l_qseq, 1);
276 for (k = 0, x = c->pos, y = 0; k < c->n_cigar; ++k) {
277 int op = cigar[k]&0xf, l = cigar[k]>>4;
278 if (op == BAM_CMATCH || op == BAM_CEQUAL || op == BAM_CDIFF) {
279 for (i = y; i < y + l; ++i)
280 bq[i] = ((state[i]&3) != 0 || state[i]>>2 != x - xb + (i - y))? 0 : q[i];
281 for (left[y] = bq[y], i = y + 1; i < y + l; ++i)
282 left[i] = bq[i] > left[i-1]? bq[i] : left[i-1];
283 for (rght[y+l-1] = bq[y+l-1], i = y + l - 2; i >= y; --i)
284 rght[i] = bq[i] > rght[i+1]? bq[i] : rght[i+1];
285 for (i = y; i < y + l; ++i)
286 bq[i] = left[i] < rght[i]? left[i] : rght[i];
287 x += l; y += l;
288 } else if (op == BAM_CSOFT_CLIP || op == BAM_CINS) y += l;
289 else if (op == BAM_CDEL) x += l;
290 }
291 for (i = 0; i < c->l_qseq; ++i) bq[i] = 64 + (qual[i] <= bq[i]? 0 : qual[i] - bq[i]); // finalize BQ
292 free(left); free(rght);
293 }
294 if (apply_baq) {
295 for (i = 0; i < c->l_qseq; ++i) qual[i] -= bq[i] - 64; // modify qual
296 bam_aux_append(b, "ZQ", 'Z', c->l_qseq + 1, bq);
297 } else bam_aux_append(b, "BQ", 'Z', c->l_qseq + 1, bq);
298 free(bq); free(s); free(r); free(q); free(state);
299 }
300 return 0;
301 }
302
303 int bam_prob_realn(bam1_t *b, const char *ref)
304 {
305 return bam_prob_realn_core(b, ref, 1);
306 }
307
308 int bam_fillmd(int argc, char *argv[])
309 {
310 int c, flt_flag, tid = -2, ret, len, is_bam_out, is_sam_in, is_uncompressed, max_nm, is_realn, capQ, baq_flag;
311 samfile_t *fp, *fpout = 0;
312 faidx_t *fai;
313 char *ref = 0, mode_w[8], mode_r[8];
314 bam1_t *b;
315
316 flt_flag = UPDATE_NM | UPDATE_MD;
317 is_bam_out = is_sam_in = is_uncompressed = is_realn = max_nm = capQ = baq_flag = 0;
318 mode_w[0] = mode_r[0] = 0;
319 strcpy(mode_r, "r"); strcpy(mode_w, "w");
320 while ((c = getopt(argc, argv, "EqreuNhbSC:n:Ad")) >= 0) {
321 switch (c) {
322 case 'r': is_realn = 1; break;
323 case 'e': flt_flag |= USE_EQUAL; break;
324 case 'd': flt_flag |= DROP_TAG; break;
325 case 'q': flt_flag |= BIN_QUAL; break;
326 case 'h': flt_flag |= HASH_QNM; break;
327 case 'N': flt_flag &= ~(UPDATE_MD|UPDATE_NM); break;
328 case 'b': is_bam_out = 1; break;
329 case 'u': is_uncompressed = is_bam_out = 1; break;
330 case 'S': is_sam_in = 1; break;
331 case 'n': max_nm = atoi(optarg); break;
332 case 'C': capQ = atoi(optarg); break;
333 case 'A': baq_flag |= 1; break;
334 case 'E': baq_flag |= 2; break;
335 default: fprintf(stderr, "[bam_fillmd] unrecognized option '-%c'\n", c); return 1;
336 }
337 }
338 if (!is_sam_in) strcat(mode_r, "b");
339 if (is_bam_out) strcat(mode_w, "b");
340 else strcat(mode_w, "h");
341 if (is_uncompressed) strcat(mode_w, "u");
342 if (optind + 1 >= argc) {
343 fprintf(stderr, "\n");
344 fprintf(stderr, "Usage: samtools fillmd [-eubrS] <aln.bam> <ref.fasta>\n\n");
345 fprintf(stderr, "Options: -e change identical bases to '='\n");
346 fprintf(stderr, " -u uncompressed BAM output (for piping)\n");
347 fprintf(stderr, " -b compressed BAM output\n");
348 fprintf(stderr, " -S the input is SAM with header\n");
349 fprintf(stderr, " -A modify the quality string\n");
350 fprintf(stderr, " -r compute the BQ tag (without -A) or cap baseQ by BAQ (with -A)\n");
351 fprintf(stderr, " -E extended BAQ for better sensitivity but lower specificity\n\n");
352 return 1;
353 }
354 fp = samopen(argv[optind], mode_r, 0);
355 if (fp == 0) return 1;
356 if (is_sam_in && (fp->header == 0 || fp->header->n_targets == 0)) {
357 fprintf(stderr, "[bam_fillmd] input SAM does not have header. Abort!\n");
358 return 1;
359 }
360 fpout = samopen("-", mode_w, fp->header);
361 fai = fai_load(argv[optind+1]);
362
363 b = bam_init1();
364 while ((ret = samread(fp, b)) >= 0) {
365 if (b->core.tid >= 0) {
366 if (tid != b->core.tid) {
367 free(ref);
368 ref = fai_fetch(fai, fp->header->target_name[b->core.tid], &len);
369 tid = b->core.tid;
370 if (ref == 0)
371 fprintf(stderr, "[bam_fillmd] fail to find sequence '%s' in the reference.\n",
372 fp->header->target_name[tid]);
373 }
374 if (is_realn) bam_prob_realn_core(b, ref, baq_flag);
375 if (capQ > 10) {
376 int q = bam_cap_mapQ(b, ref, capQ);
377 if (b->core.qual > q) b->core.qual = q;
378 }
379 if (ref) bam_fillmd1_core(b, ref, flt_flag, max_nm);
380 }
381 samwrite(fpout, b);
382 }
383 bam_destroy1(b);
384
385 free(ref);
386 fai_destroy(fai);
387 samclose(fp); samclose(fpout);
388 return 0;
389 }