|
0
|
1 #include <stdio.h>
|
|
|
2 #include <stdlib.h>
|
|
|
3 #include <unistd.h>
|
|
|
4 #include <stdint.h>
|
|
|
5 #include <math.h>
|
|
|
6 #include <zlib.h>
|
|
|
7 #include "bam.h"
|
|
|
8 #include "errmod.h"
|
|
|
9
|
|
|
10 #include "kseq.h"
|
|
|
11 KSTREAM_INIT(gzFile, gzread, 16384)
|
|
|
12
|
|
|
13 #define MAX_VARS 256
|
|
|
14 #define FLIP_PENALTY 2
|
|
|
15 #define FLIP_THRES 4
|
|
|
16 #define MASK_THRES 3
|
|
|
17
|
|
|
18 #define FLAG_FIX_CHIMERA 0x1
|
|
|
19 #define FLAG_LIST_EXCL 0x4
|
|
|
20 #define FLAG_DROP_AMBI 0x8
|
|
|
21
|
|
|
22 typedef struct {
|
|
|
23 // configurations, initialized in the main function
|
|
|
24 int flag, k, min_baseQ, min_varLOD, max_depth;
|
|
|
25 // other global variables
|
|
|
26 int vpos_shift;
|
|
|
27 bamFile fp;
|
|
|
28 char *pre;
|
|
|
29 bamFile out[3];
|
|
|
30 // alignment queue
|
|
|
31 int n, m;
|
|
|
32 bam1_t **b;
|
|
|
33 } phaseg_t;
|
|
|
34
|
|
|
35 typedef struct {
|
|
|
36 int8_t seq[MAX_VARS]; // TODO: change to dynamic memory allocation!
|
|
|
37 int vpos, beg, end;
|
|
|
38 uint32_t vlen:16, single:1, flip:1, phase:1, phased:1, ambig:1;
|
|
|
39 uint32_t in:16, out:16; // in-phase and out-phase
|
|
|
40 } frag_t, *frag_p;
|
|
|
41
|
|
|
42 #define rseq_lt(a,b) ((a)->vpos < (b)->vpos)
|
|
|
43
|
|
|
44 #include "khash.h"
|
|
|
45 KHASH_SET_INIT_INT64(set64)
|
|
|
46 KHASH_MAP_INIT_INT64(64, frag_t)
|
|
|
47
|
|
|
48 typedef khash_t(64) nseq_t;
|
|
|
49
|
|
|
50 #include "ksort.h"
|
|
|
51 KSORT_INIT(rseq, frag_p, rseq_lt)
|
|
|
52
|
|
|
53 static char nt16_nt4_table[] = { 4, 0, 1, 4, 2, 4, 4, 4, 3, 4, 4, 4, 4, 4, 4, 4 };
|
|
|
54
|
|
|
55 static inline uint64_t X31_hash_string(const char *s)
|
|
|
56 {
|
|
|
57 uint64_t h = *s;
|
|
|
58 if (h) for (++s ; *s; ++s) h = (h << 5) - h + *s;
|
|
|
59 return h;
|
|
|
60 }
|
|
|
61
|
|
|
62 static void count1(int l, const uint8_t *seq, int *cnt)
|
|
|
63 {
|
|
|
64 int i, j, n_ambi;
|
|
|
65 uint32_t z, x;
|
|
|
66 if (seq[l-1] == 0) return; // do nothing is the last base is ambiguous
|
|
|
67 for (i = n_ambi = 0; i < l; ++i) // collect ambiguous bases
|
|
|
68 if (seq[i] == 0) ++n_ambi;
|
|
|
69 if (l - n_ambi <= 1) return; // only one SNP
|
|
|
70 for (x = 0; x < 1u<<n_ambi; ++x) { // count
|
|
|
71 for (i = j = 0, z = 0; i < l; ++i) {
|
|
|
72 int c;
|
|
|
73 if (seq[i]) c = seq[i] - 1;
|
|
|
74 else {
|
|
|
75 c = x>>j&1;
|
|
|
76 ++j;
|
|
|
77 }
|
|
|
78 z = z<<1 | c;
|
|
|
79 }
|
|
|
80 ++cnt[z];
|
|
|
81 }
|
|
|
82 }
|
|
|
83
|
|
|
84 static int **count_all(int l, int vpos, nseq_t *hash)
|
|
|
85 {
|
|
|
86 khint_t k;
|
|
|
87 int i, j, **cnt;
|
|
|
88 uint8_t *seq;
|
|
|
89 seq = calloc(l, 1);
|
|
|
90 cnt = calloc(vpos, sizeof(void*));
|
|
|
91 for (i = 0; i < vpos; ++i) cnt[i] = calloc(1<<l, sizeof(int));
|
|
|
92 for (k = 0; k < kh_end(hash); ++k) {
|
|
|
93 if (kh_exist(hash, k)) {
|
|
|
94 frag_t *f = &kh_val(hash, k);
|
|
|
95 if (f->vpos >= vpos || f->single) continue; // out of region; or singleton
|
|
|
96 if (f->vlen == 1) { // such reads should be flagged as deleted previously if everything is right
|
|
|
97 f->single = 1;
|
|
|
98 continue;
|
|
|
99 }
|
|
|
100 for (j = 1; j < f->vlen; ++j) {
|
|
|
101 for (i = 0; i < l; ++i)
|
|
|
102 seq[i] = j < l - 1 - i? 0 : f->seq[j - (l - 1 - i)];
|
|
|
103 count1(l, seq, cnt[f->vpos + j]);
|
|
|
104 }
|
|
|
105 }
|
|
|
106 }
|
|
|
107 free(seq);
|
|
|
108 return cnt;
|
|
|
109 }
|
|
|
110
|
|
|
111 // phasing
|
|
|
112 static int8_t *dynaprog(int l, int vpos, int **w)
|
|
|
113 {
|
|
|
114 int *f[2], *curr, *prev, max, i;
|
|
|
115 int8_t **b, *h = 0;
|
|
|
116 uint32_t x, z = 1u<<(l-1), mask = (1u<<l) - 1;
|
|
|
117 f[0] = calloc(z, sizeof(int));
|
|
|
118 f[1] = calloc(z, sizeof(int));
|
|
|
119 b = calloc(vpos, sizeof(void*));
|
|
|
120 prev = f[0]; curr = f[1];
|
|
|
121 // fill the backtrack matrix
|
|
|
122 for (i = 0; i < vpos; ++i) {
|
|
|
123 int *wi = w[i], *tmp;
|
|
|
124 int8_t *bi;
|
|
|
125 bi = b[i] = calloc(z, 1);
|
|
|
126 /* In the following, x is the current state, which is the
|
|
|
127 * lexicographically smaller local haplotype. xc is the complement of
|
|
|
128 * x, or the larger local haplotype; y0 and y1 are the two predecessors
|
|
|
129 * of x. */
|
|
|
130 for (x = 0; x < z; ++x) { // x0 is the smaller
|
|
|
131 uint32_t y0, y1, xc;
|
|
|
132 int c0, c1;
|
|
|
133 xc = ~x&mask; y0 = x>>1; y1 = xc>>1;
|
|
|
134 c0 = prev[y0] + wi[x] + wi[xc];
|
|
|
135 c1 = prev[y1] + wi[x] + wi[xc];
|
|
|
136 if (c0 > c1) bi[x] = 0, curr[x] = c0;
|
|
|
137 else bi[x] = 1, curr[x] = c1;
|
|
|
138 }
|
|
|
139 tmp = prev; prev = curr; curr = tmp; // swap
|
|
|
140 }
|
|
|
141 { // backtrack
|
|
|
142 uint32_t max_x = 0;
|
|
|
143 int which = 0;
|
|
|
144 h = calloc(vpos, 1);
|
|
|
145 for (x = 0, max = 0, max_x = 0; x < z; ++x)
|
|
|
146 if (prev[x] > max) max = prev[x], max_x = x;
|
|
|
147 for (i = vpos - 1, x = max_x; i >= 0; --i) {
|
|
|
148 h[i] = which? (~x&1) : (x&1);
|
|
|
149 which = b[i][x]? !which : which;
|
|
|
150 x = b[i][x]? (~x&mask)>>1 : x>>1;
|
|
|
151 }
|
|
|
152 }
|
|
|
153 // free
|
|
|
154 for (i = 0; i < vpos; ++i) free(b[i]);
|
|
|
155 free(f[0]); free(f[1]); free(b);
|
|
|
156 return h;
|
|
|
157 }
|
|
|
158
|
|
|
159 // phase each fragment
|
|
|
160 static uint64_t *fragphase(int vpos, const int8_t *path, nseq_t *hash, int flip)
|
|
|
161 {
|
|
|
162 khint_t k;
|
|
|
163 uint64_t *pcnt;
|
|
|
164 uint32_t *left, *rght, max;
|
|
|
165 left = rght = 0; max = 0;
|
|
|
166 pcnt = calloc(vpos, 8);
|
|
|
167 for (k = 0; k < kh_end(hash); ++k) {
|
|
|
168 if (kh_exist(hash, k)) {
|
|
|
169 int i, c[2];
|
|
|
170 frag_t *f = &kh_val(hash, k);
|
|
|
171 if (f->vpos >= vpos) continue;
|
|
|
172 // get the phase
|
|
|
173 c[0] = c[1] = 0;
|
|
|
174 for (i = 0; i < f->vlen; ++i) {
|
|
|
175 if (f->seq[i] == 0) continue;
|
|
|
176 ++c[f->seq[i] == path[f->vpos + i] + 1? 0 : 1];
|
|
|
177 }
|
|
|
178 f->phase = c[0] > c[1]? 0 : 1;
|
|
|
179 f->in = c[f->phase]; f->out = c[1 - f->phase];
|
|
|
180 f->phased = f->in == f->out? 0 : 1;
|
|
|
181 f->ambig = (f->in && f->out && f->out < 3 && f->in <= f->out + 1)? 1 : 0;
|
|
|
182 // fix chimera
|
|
|
183 f->flip = 0;
|
|
|
184 if (flip && c[0] >= 3 && c[1] >= 3) {
|
|
|
185 int sum[2], m, mi, md;
|
|
|
186 if (f->vlen > max) { // enlarge the array
|
|
|
187 max = f->vlen;
|
|
|
188 kroundup32(max);
|
|
|
189 left = realloc(left, max * 4);
|
|
|
190 rght = realloc(rght, max * 4);
|
|
|
191 }
|
|
|
192 for (i = 0, sum[0] = sum[1] = 0; i < f->vlen; ++i) { // get left counts
|
|
|
193 if (f->seq[i]) {
|
|
|
194 int c = f->phase? 2 - f->seq[i] : f->seq[i] - 1;
|
|
|
195 ++sum[c == path[f->vpos + i]? 0 : 1];
|
|
|
196 }
|
|
|
197 left[i] = sum[1]<<16 | sum[0];
|
|
|
198 }
|
|
|
199 for (i = f->vlen - 1, sum[0] = sum[1] = 0; i >= 0; --i) { // get right counts
|
|
|
200 if (f->seq[i]) {
|
|
|
201 int c = f->phase? 2 - f->seq[i] : f->seq[i] - 1;
|
|
|
202 ++sum[c == path[f->vpos + i]? 0 : 1];
|
|
|
203 }
|
|
|
204 rght[i] = sum[1]<<16 | sum[0];
|
|
|
205 }
|
|
|
206 // find the best flip point
|
|
|
207 for (i = m = 0, mi = -1, md = -1; i < f->vlen - 1; ++i) {
|
|
|
208 int a[2];
|
|
|
209 a[0] = (left[i]&0xffff) + (rght[i+1]>>16&0xffff) - (rght[i+1]&0xffff) * FLIP_PENALTY;
|
|
|
210 a[1] = (left[i]>>16&0xffff) + (rght[i+1]&0xffff) - (rght[i+1]>>16&0xffff) * FLIP_PENALTY;
|
|
|
211 if (a[0] > a[1]) {
|
|
|
212 if (a[0] > m) m = a[0], md = 0, mi = i;
|
|
|
213 } else {
|
|
|
214 if (a[1] > m) m = a[1], md = 1, mi = i;
|
|
|
215 }
|
|
|
216 }
|
|
|
217 if (m - c[0] >= FLIP_THRES && m - c[1] >= FLIP_THRES) { // then flip
|
|
|
218 f->flip = 1;
|
|
|
219 if (md == 0) { // flip the tail
|
|
|
220 for (i = mi + 1; i < f->vlen; ++i)
|
|
|
221 if (f->seq[i] == 1) f->seq[i] = 2;
|
|
|
222 else if (f->seq[i] == 2) f->seq[i] = 1;
|
|
|
223 } else { // flip the head
|
|
|
224 for (i = 0; i <= mi; ++i)
|
|
|
225 if (f->seq[i] == 1) f->seq[i] = 2;
|
|
|
226 else if (f->seq[i] == 2) f->seq[i] = 1;
|
|
|
227 }
|
|
|
228 }
|
|
|
229 }
|
|
|
230 // update pcnt[]
|
|
|
231 if (!f->single) {
|
|
|
232 for (i = 0; i < f->vlen; ++i) {
|
|
|
233 int c;
|
|
|
234 if (f->seq[i] == 0) continue;
|
|
|
235 c = f->phase? 2 - f->seq[i] : f->seq[i] - 1;
|
|
|
236 if (c == path[f->vpos + i]) {
|
|
|
237 if (f->phase == 0) ++pcnt[f->vpos + i];
|
|
|
238 else pcnt[f->vpos + i] += 1ull<<32;
|
|
|
239 } else {
|
|
|
240 if (f->phase == 0) pcnt[f->vpos + i] += 1<<16;
|
|
|
241 else pcnt[f->vpos + i] += 1ull<<48;
|
|
|
242 }
|
|
|
243 }
|
|
|
244 }
|
|
|
245 }
|
|
|
246 }
|
|
|
247 free(left); free(rght);
|
|
|
248 return pcnt;
|
|
|
249 }
|
|
|
250
|
|
|
251 static uint64_t *genmask(int vpos, const uint64_t *pcnt, int *_n)
|
|
|
252 {
|
|
|
253 int i, max = 0, max_i = -1, m = 0, n = 0, beg = 0, score = 0;
|
|
|
254 uint64_t *list = 0;
|
|
|
255 for (i = 0; i < vpos; ++i) {
|
|
|
256 uint64_t x = pcnt[i];
|
|
|
257 int c[4], pre = score, s;
|
|
|
258 c[0] = x&0xffff; c[1] = x>>16&0xffff; c[2] = x>>32&0xffff; c[3] = x>>48&0xffff;
|
|
|
259 s = (c[1] + c[3] == 0)? -(c[0] + c[2]) : (c[1] + c[3] - 1);
|
|
|
260 if (c[3] > c[2]) s += c[3] - c[2];
|
|
|
261 if (c[1] > c[0]) s += c[1] - c[0];
|
|
|
262 score += s;
|
|
|
263 if (score < 0) score = 0;
|
|
|
264 if (pre == 0 && score > 0) beg = i; // change from zero to non-zero
|
|
|
265 if ((i == vpos - 1 || score == 0) && max >= MASK_THRES) {
|
|
|
266 if (n == m) {
|
|
|
267 m = m? m<<1 : 4;
|
|
|
268 list = realloc(list, m * 8);
|
|
|
269 }
|
|
|
270 list[n++] = (uint64_t)beg<<32 | max_i;
|
|
|
271 i = max_i; // reset i to max_i
|
|
|
272 score = 0;
|
|
|
273 } else if (score > max) max = score, max_i = i;
|
|
|
274 if (score == 0) max = 0;
|
|
|
275 }
|
|
|
276 *_n = n;
|
|
|
277 return list;
|
|
|
278 }
|
|
|
279
|
|
|
280 // trim heading and tailing ambiguous bases; mark deleted and remove sequence
|
|
|
281 static int clean_seqs(int vpos, nseq_t *hash)
|
|
|
282 {
|
|
|
283 khint_t k;
|
|
|
284 int ret = 0;
|
|
|
285 for (k = 0; k < kh_end(hash); ++k) {
|
|
|
286 if (kh_exist(hash, k)) {
|
|
|
287 frag_t *f = &kh_val(hash, k);
|
|
|
288 int beg, end, i;
|
|
|
289 if (f->vpos >= vpos) {
|
|
|
290 ret = 1;
|
|
|
291 continue;
|
|
|
292 }
|
|
|
293 for (i = 0; i < f->vlen; ++i)
|
|
|
294 if (f->seq[i] != 0) break;
|
|
|
295 beg = i;
|
|
|
296 for (i = f->vlen - 1; i >= 0; --i)
|
|
|
297 if (f->seq[i] != 0) break;
|
|
|
298 end = i + 1;
|
|
|
299 if (end - beg <= 0) kh_del(64, hash, k);
|
|
|
300 else {
|
|
|
301 if (beg != 0) memmove(f->seq, f->seq + beg, end - beg);
|
|
|
302 f->vpos += beg; f->vlen = end - beg;
|
|
|
303 f->single = f->vlen == 1? 1 : 0;
|
|
|
304 }
|
|
|
305 }
|
|
|
306 }
|
|
|
307 return ret;
|
|
|
308 }
|
|
|
309
|
|
|
310 static void dump_aln(phaseg_t *g, int min_pos, const nseq_t *hash)
|
|
|
311 {
|
|
|
312 int i, is_flip, drop_ambi;
|
|
|
313 drop_ambi = g->flag & FLAG_DROP_AMBI;
|
|
|
314 is_flip = (drand48() < 0.5);
|
|
|
315 for (i = 0; i < g->n; ++i) {
|
|
|
316 int end, which;
|
|
|
317 uint64_t key;
|
|
|
318 khint_t k;
|
|
|
319 bam1_t *b = g->b[i];
|
|
|
320 key = X31_hash_string(bam1_qname(b));
|
|
|
321 end = bam_calend(&b->core, bam1_cigar(b));
|
|
|
322 if (end > min_pos) break;
|
|
|
323 k = kh_get(64, hash, key);
|
|
|
324 if (k == kh_end(hash)) which = 3;
|
|
|
325 else {
|
|
|
326 frag_t *f = &kh_val(hash, k);
|
|
|
327 if (f->ambig) which = drop_ambi? 2 : 3;
|
|
|
328 else if (f->phased && f->flip) which = 2;
|
|
|
329 else if (f->phased == 0) which = 3;
|
|
|
330 else { // phased and not flipped
|
|
|
331 char c = 'Y';
|
|
|
332 which = f->phase;
|
|
|
333 bam_aux_append(b, "ZP", 'A', 1, (uint8_t*)&c);
|
|
|
334 }
|
|
|
335 if (which < 2 && is_flip) which = 1 - which; // increase the randomness
|
|
|
336 }
|
|
|
337 if (which == 3) which = (drand48() < 0.5);
|
|
|
338 bam_write1(g->out[which], b);
|
|
|
339 bam_destroy1(b);
|
|
|
340 g->b[i] = 0;
|
|
|
341 }
|
|
|
342 memmove(g->b, g->b + i, (g->n - i) * sizeof(void*));
|
|
|
343 g->n -= i;
|
|
|
344 }
|
|
|
345
|
|
|
346 static int phase(phaseg_t *g, const char *chr, int vpos, uint64_t *cns, nseq_t *hash)
|
|
|
347 {
|
|
|
348 int i, j, n_seqs = kh_size(hash), n_masked = 0, min_pos;
|
|
|
349 khint_t k;
|
|
|
350 frag_t **seqs;
|
|
|
351 int8_t *path, *sitemask;
|
|
|
352 uint64_t *pcnt, *regmask;
|
|
|
353
|
|
|
354 if (vpos == 0) return 0;
|
|
|
355 i = clean_seqs(vpos, hash); // i is true if hash has an element with its vpos >= vpos
|
|
|
356 min_pos = i? cns[vpos]>>32 : 0x7fffffff;
|
|
|
357 if (vpos == 1) {
|
|
|
358 printf("PS\t%s\t%d\t%d\n", chr, (int)(cns[0]>>32) + 1, (int)(cns[0]>>32) + 1);
|
|
|
359 printf("M0\t%s\t%d\t%d\t%c\t%c\t%d\t0\t0\t0\t0\n//\n", chr, (int)(cns[0]>>32) + 1, (int)(cns[0]>>32) + 1,
|
|
|
360 "ACGTX"[cns[0]&3], "ACGTX"[cns[0]>>16&3], g->vpos_shift + 1);
|
|
|
361 for (k = 0; k < kh_end(hash); ++k) {
|
|
|
362 if (kh_exist(hash, k)) {
|
|
|
363 frag_t *f = &kh_val(hash, k);
|
|
|
364 if (f->vpos) continue;
|
|
|
365 f->flip = 0;
|
|
|
366 if (f->seq[0] == 0) f->phased = 0;
|
|
|
367 else f->phased = 1, f->phase = f->seq[0] - 1;
|
|
|
368 }
|
|
|
369 }
|
|
|
370 dump_aln(g, min_pos, hash);
|
|
|
371 ++g->vpos_shift;
|
|
|
372 return 1;
|
|
|
373 }
|
|
|
374 { // phase
|
|
|
375 int **cnt;
|
|
|
376 uint64_t *mask;
|
|
|
377 printf("PS\t%s\t%d\t%d\n", chr, (int)(cns[0]>>32) + 1, (int)(cns[vpos-1]>>32) + 1);
|
|
|
378 sitemask = calloc(vpos, 1);
|
|
|
379 cnt = count_all(g->k, vpos, hash);
|
|
|
380 path = dynaprog(g->k, vpos, cnt);
|
|
|
381 for (i = 0; i < vpos; ++i) free(cnt[i]);
|
|
|
382 free(cnt);
|
|
|
383 pcnt = fragphase(vpos, path, hash, 0); // do not fix chimeras when masking
|
|
|
384 mask = genmask(vpos, pcnt, &n_masked);
|
|
|
385 regmask = calloc(n_masked, 8);
|
|
|
386 for (i = 0; i < n_masked; ++i) {
|
|
|
387 regmask[i] = cns[mask[i]>>32]>>32<<32 | cns[(uint32_t)mask[i]]>>32;
|
|
|
388 for (j = mask[i]>>32; j <= (int32_t)mask[i]; ++j)
|
|
|
389 sitemask[j] = 1;
|
|
|
390 }
|
|
|
391 free(mask);
|
|
|
392 if (g->flag & FLAG_FIX_CHIMERA) {
|
|
|
393 free(pcnt);
|
|
|
394 pcnt = fragphase(vpos, path, hash, 1);
|
|
|
395 }
|
|
|
396 }
|
|
|
397 for (i = 0; i < n_masked; ++i)
|
|
|
398 printf("FL\t%s\t%d\t%d\n", chr, (int)(regmask[i]>>32) + 1, (int)regmask[i] + 1);
|
|
|
399 for (i = 0; i < vpos; ++i) {
|
|
|
400 uint64_t x = pcnt[i];
|
|
|
401 int8_t c[2];
|
|
|
402 c[0] = (cns[i]&0xffff)>>2 == 0? 4 : (cns[i]&3);
|
|
|
403 c[1] = (cns[i]>>16&0xffff)>>2 == 0? 4 : (cns[i]>>16&3);
|
|
|
404 printf("M%d\t%s\t%d\t%d\t%c\t%c\t%d\t%d\t%d\t%d\t%d\n", sitemask[i]+1, chr, (int)(cns[0]>>32) + 1, (int)(cns[i]>>32) + 1, "ACGTX"[c[path[i]]], "ACGTX"[c[1-path[i]]],
|
|
|
405 i + g->vpos_shift + 1, (int)(x&0xffff), (int)(x>>16&0xffff), (int)(x>>32&0xffff), (int)(x>>48&0xffff));
|
|
|
406 }
|
|
|
407 free(path); free(pcnt); free(regmask); free(sitemask);
|
|
|
408 seqs = calloc(n_seqs, sizeof(void*));
|
|
|
409 for (k = 0, i = 0; k < kh_end(hash); ++k)
|
|
|
410 if (kh_exist(hash, k) && kh_val(hash, k).vpos < vpos && !kh_val(hash, k).single)
|
|
|
411 seqs[i++] = &kh_val(hash, k);
|
|
|
412 n_seqs = i;
|
|
|
413 ks_introsort_rseq(n_seqs, seqs);
|
|
|
414 for (i = 0; i < n_seqs; ++i) {
|
|
|
415 frag_t *f = seqs[i];
|
|
|
416 printf("EV\t0\t%s\t%d\t40\t%dM\t*\t0\t0\t", chr, f->vpos + 1 + g->vpos_shift, f->vlen);
|
|
|
417 for (j = 0; j < f->vlen; ++j) {
|
|
|
418 uint32_t c = cns[f->vpos + j];
|
|
|
419 if (f->seq[j] == 0) putchar('N');
|
|
|
420 else putchar("ACGT"[f->seq[j] == 1? (c&3) : (c>>16&3)]);
|
|
|
421 }
|
|
|
422 printf("\t*\tYP:i:%d\tYF:i:%d\tYI:i:%d\tYO:i:%d\tYS:i:%d\n", f->phase, f->flip, f->in, f->out, f->beg+1);
|
|
|
423 }
|
|
|
424 free(seqs);
|
|
|
425 printf("//\n");
|
|
|
426 fflush(stdout);
|
|
|
427 g->vpos_shift += vpos;
|
|
|
428 dump_aln(g, min_pos, hash);
|
|
|
429 return vpos;
|
|
|
430 }
|
|
|
431
|
|
|
432 static void update_vpos(int vpos, nseq_t *hash)
|
|
|
433 {
|
|
|
434 khint_t k;
|
|
|
435 for (k = 0; k < kh_end(hash); ++k) {
|
|
|
436 if (kh_exist(hash, k)) {
|
|
|
437 frag_t *f = &kh_val(hash, k);
|
|
|
438 if (f->vpos < vpos) kh_del(64, hash, k); // TODO: if frag_t::seq is allocated dynamically, free it
|
|
|
439 else f->vpos -= vpos;
|
|
|
440 }
|
|
|
441 }
|
|
|
442 }
|
|
|
443
|
|
|
444 static nseq_t *shrink_hash(nseq_t *hash) // TODO: to implement
|
|
|
445 {
|
|
|
446 return hash;
|
|
|
447 }
|
|
|
448
|
|
|
449 static int readaln(void *data, bam1_t *b)
|
|
|
450 {
|
|
|
451 phaseg_t *g = (phaseg_t*)data;
|
|
|
452 int ret;
|
|
|
453 ret = bam_read1(g->fp, b);
|
|
|
454 if (ret < 0) return ret;
|
|
|
455 if (!(b->core.flag & (BAM_FUNMAP|BAM_FSECONDARY|BAM_FQCFAIL|BAM_FDUP)) && g->pre) {
|
|
|
456 if (g->n == g->m) {
|
|
|
457 g->m = g->m? g->m<<1 : 16;
|
|
|
458 g->b = realloc(g->b, g->m * sizeof(void*));
|
|
|
459 }
|
|
|
460 g->b[g->n++] = bam_dup1(b);
|
|
|
461 }
|
|
|
462 return ret;
|
|
|
463 }
|
|
|
464
|
|
|
465 static khash_t(set64) *loadpos(const char *fn, bam_header_t *h)
|
|
|
466 {
|
|
|
467 gzFile fp;
|
|
|
468 kstream_t *ks;
|
|
|
469 int ret, dret;
|
|
|
470 kstring_t *str;
|
|
|
471 khash_t(set64) *hash;
|
|
|
472
|
|
|
473 hash = kh_init(set64);
|
|
|
474 str = calloc(1, sizeof(kstring_t));
|
|
|
475 fp = strcmp(fn, "-")? gzopen(fn, "r") : gzdopen(fileno(stdin), "r");
|
|
|
476 ks = ks_init(fp);
|
|
|
477 while (ks_getuntil(ks, 0, str, &dret) >= 0) {
|
|
|
478 int tid = bam_get_tid(h, str->s);
|
|
|
479 if (tid >= 0 && dret != '\n') {
|
|
|
480 if (ks_getuntil(ks, 0, str, &dret) >= 0) {
|
|
|
481 uint64_t x = (uint64_t)tid<<32 | (atoi(str->s) - 1);
|
|
|
482 kh_put(set64, hash, x, &ret);
|
|
|
483 } else break;
|
|
|
484 }
|
|
|
485 if (dret != '\n') while ((dret = ks_getc(ks)) > 0 && dret != '\n');
|
|
|
486 if (dret < 0) break;
|
|
|
487 }
|
|
|
488 ks_destroy(ks);
|
|
|
489 gzclose(fp);
|
|
|
490 free(str->s); free(str);
|
|
|
491 return hash;
|
|
|
492 }
|
|
|
493
|
|
|
494 static int gl2cns(float q[16])
|
|
|
495 {
|
|
|
496 int i, j, min_ij;
|
|
|
497 float min, min2;
|
|
|
498 min = min2 = 1e30; min_ij = -1;
|
|
|
499 for (i = 0; i < 4; ++i) {
|
|
|
500 for (j = i; j < 4; ++j) {
|
|
|
501 if (q[i<<2|j] < min) min_ij = i<<2|j, min2 = min, min = q[i<<2|j];
|
|
|
502 else if (q[i<<2|j] < min2) min2 = q[i<<2|j];
|
|
|
503 }
|
|
|
504 }
|
|
|
505 return (min_ij>>2&3) == (min_ij&3)? 0 : 1<<18 | (min_ij>>2&3)<<16 | (min_ij&3) | (int)(min2 - min + .499) << 2;
|
|
|
506 }
|
|
|
507
|
|
|
508 int main_phase(int argc, char *argv[])
|
|
|
509 {
|
|
|
510 extern void bam_init_header_hash(bam_header_t *header);
|
|
|
511 int c, tid, pos, vpos = 0, n, lasttid = -1, max_vpos = 0;
|
|
|
512 const bam_pileup1_t *plp;
|
|
|
513 bam_plp_t iter;
|
|
|
514 bam_header_t *h;
|
|
|
515 nseq_t *seqs;
|
|
|
516 uint64_t *cns = 0;
|
|
|
517 phaseg_t g;
|
|
|
518 char *fn_list = 0;
|
|
|
519 khash_t(set64) *set = 0;
|
|
|
520 errmod_t *em;
|
|
|
521 uint16_t *bases;
|
|
|
522
|
|
|
523 memset(&g, 0, sizeof(phaseg_t));
|
|
|
524 g.flag = FLAG_FIX_CHIMERA;
|
|
|
525 g.min_varLOD = 37; g.k = 13; g.min_baseQ = 13; g.max_depth = 256;
|
|
|
526 while ((c = getopt(argc, argv, "Q:eFq:k:b:l:D:A:")) >= 0) {
|
|
|
527 switch (c) {
|
|
|
528 case 'D': g.max_depth = atoi(optarg); break;
|
|
|
529 case 'q': g.min_varLOD = atoi(optarg); break;
|
|
|
530 case 'Q': g.min_baseQ = atoi(optarg); break;
|
|
|
531 case 'k': g.k = atoi(optarg); break;
|
|
|
532 case 'F': g.flag &= ~FLAG_FIX_CHIMERA; break;
|
|
|
533 case 'e': g.flag |= FLAG_LIST_EXCL; break;
|
|
|
534 case 'A': g.flag |= FLAG_DROP_AMBI; break;
|
|
|
535 case 'b': g.pre = strdup(optarg); break;
|
|
|
536 case 'l': fn_list = strdup(optarg); break;
|
|
|
537 }
|
|
|
538 }
|
|
|
539 if (argc == optind) {
|
|
|
540 fprintf(stderr, "\n");
|
|
|
541 fprintf(stderr, "Usage: samtools phase [options] <in.bam>\n\n");
|
|
|
542 fprintf(stderr, "Options: -k INT block length [%d]\n", g.k);
|
|
|
543 fprintf(stderr, " -b STR prefix of BAMs to output [null]\n");
|
|
|
544 fprintf(stderr, " -q INT min het phred-LOD [%d]\n", g.min_varLOD);
|
|
|
545 fprintf(stderr, " -Q INT min base quality in het calling [%d]\n", g.min_baseQ);
|
|
|
546 fprintf(stderr, " -D INT max read depth [%d]\n", g.max_depth);
|
|
|
547 // fprintf(stderr, " -l FILE list of sites to phase [null]\n");
|
|
|
548 fprintf(stderr, " -F do not attempt to fix chimeras\n");
|
|
|
549 fprintf(stderr, " -A drop reads with ambiguous phase\n");
|
|
|
550 // fprintf(stderr, " -e do not discover SNPs (effective with -l)\n");
|
|
|
551 fprintf(stderr, "\n");
|
|
|
552 return 1;
|
|
|
553 }
|
|
|
554 g.fp = strcmp(argv[optind], "-")? bam_open(argv[optind], "r") : bam_dopen(fileno(stdin), "r");
|
|
|
555 h = bam_header_read(g.fp);
|
|
|
556 if (fn_list) { // read the list of sites to phase
|
|
|
557 bam_init_header_hash(h);
|
|
|
558 set = loadpos(fn_list, h);
|
|
|
559 free(fn_list);
|
|
|
560 } else g.flag &= ~FLAG_LIST_EXCL;
|
|
|
561 if (g.pre) { // open BAMs to write
|
|
|
562 char *s = malloc(strlen(g.pre) + 20);
|
|
|
563 strcpy(s, g.pre); strcat(s, ".0.bam"); g.out[0] = bam_open(s, "w");
|
|
|
564 strcpy(s, g.pre); strcat(s, ".1.bam"); g.out[1] = bam_open(s, "w");
|
|
|
565 strcpy(s, g.pre); strcat(s, ".chimera.bam"); g.out[2] = bam_open(s, "w");
|
|
|
566 for (c = 0; c <= 2; ++c) bam_header_write(g.out[c], h);
|
|
|
567 free(s);
|
|
|
568 }
|
|
|
569
|
|
|
570 iter = bam_plp_init(readaln, &g);
|
|
|
571 g.vpos_shift = 0;
|
|
|
572 seqs = kh_init(64);
|
|
|
573 em = errmod_init(1. - 0.83);
|
|
|
574 bases = calloc(g.max_depth, 2);
|
|
|
575 printf("CC\n");
|
|
|
576 printf("CC\tDescriptions:\nCC\n");
|
|
|
577 printf("CC\t CC comments\n");
|
|
|
578 printf("CC\t PS start of a phase set\n");
|
|
|
579 printf("CC\t FL filtered region\n");
|
|
|
580 printf("CC\t M[012] markers; 0 for singletons, 1 for phased and 2 for filtered\n");
|
|
|
581 printf("CC\t EV supporting reads; SAM format\n");
|
|
|
582 printf("CC\t // end of a phase set\nCC\n");
|
|
|
583 printf("CC\tFormats of PS, FL and M[012] lines (1-based coordinates):\nCC\n");
|
|
|
584 printf("CC\t PS chr phaseSetStart phaseSetEnd\n");
|
|
|
585 printf("CC\t FL chr filterStart filterEnd\n");
|
|
|
586 printf("CC\t M? chr PS pos allele0 allele1 hetIndex #supports0 #errors0 #supp1 #err1\n");
|
|
|
587 printf("CC\nCC\n");
|
|
|
588 fflush(stdout);
|
|
|
589 while ((plp = bam_plp_auto(iter, &tid, &pos, &n)) != 0) {
|
|
|
590 int i, k, c, tmp, dophase = 1, in_set = 0;
|
|
|
591 float q[16];
|
|
|
592 if (tid < 0) break;
|
|
|
593 if (tid != lasttid) { // change of chromosome
|
|
|
594 g.vpos_shift = 0;
|
|
|
595 if (lasttid >= 0) {
|
|
|
596 seqs = shrink_hash(seqs);
|
|
|
597 phase(&g, h->target_name[lasttid], vpos, cns, seqs);
|
|
|
598 update_vpos(0x7fffffff, seqs);
|
|
|
599 }
|
|
|
600 lasttid = tid;
|
|
|
601 vpos = 0;
|
|
|
602 }
|
|
|
603 if (set && kh_get(set64, set, (uint64_t)tid<<32 | pos) != kh_end(set)) in_set = 1;
|
|
|
604 if (n > g.max_depth) continue; // do not proceed if the depth is too high
|
|
|
605 // fill the bases array and check if there is a variant
|
|
|
606 for (i = k = 0; i < n; ++i) {
|
|
|
607 const bam_pileup1_t *p = plp + i;
|
|
|
608 uint8_t *seq;
|
|
|
609 int q, baseQ, b;
|
|
|
610 if (p->is_del || p->is_refskip) continue;
|
|
|
611 baseQ = bam1_qual(p->b)[p->qpos];
|
|
|
612 if (baseQ < g.min_baseQ) continue;
|
|
|
613 seq = bam1_seq(p->b);
|
|
|
614 b = bam_nt16_nt4_table[bam1_seqi(seq, p->qpos)];
|
|
|
615 if (b > 3) continue;
|
|
|
616 q = baseQ < p->b->core.qual? baseQ : p->b->core.qual;
|
|
|
617 if (q < 4) q = 4;
|
|
|
618 if (q > 63) q = 63;
|
|
|
619 bases[k++] = q<<5 | (int)bam1_strand(p->b)<<4 | b;
|
|
|
620 }
|
|
|
621 if (k == 0) continue;
|
|
|
622 errmod_cal(em, k, 4, bases, q); // compute genotype likelihood
|
|
|
623 c = gl2cns(q); // get the consensus
|
|
|
624 // tell if to proceed
|
|
|
625 if (set && (g.flag&FLAG_LIST_EXCL) && !in_set) continue; // not in the list
|
|
|
626 if (!in_set && (c&0xffff)>>2 < g.min_varLOD) continue; // not a variant
|
|
|
627 // add the variant
|
|
|
628 if (vpos == max_vpos) {
|
|
|
629 max_vpos = max_vpos? max_vpos<<1 : 128;
|
|
|
630 cns = realloc(cns, max_vpos * 8);
|
|
|
631 }
|
|
|
632 cns[vpos] = (uint64_t)pos<<32 | c;
|
|
|
633 for (i = 0; i < n; ++i) {
|
|
|
634 const bam_pileup1_t *p = plp + i;
|
|
|
635 uint64_t key;
|
|
|
636 khint_t k;
|
|
|
637 uint8_t *seq = bam1_seq(p->b);
|
|
|
638 frag_t *f;
|
|
|
639 if (p->is_del || p->is_refskip) continue;
|
|
|
640 if (p->b->core.qual == 0) continue;
|
|
|
641 // get the base code
|
|
|
642 c = nt16_nt4_table[(int)bam1_seqi(seq, p->qpos)];
|
|
|
643 if (c == (cns[vpos]&3)) c = 1;
|
|
|
644 else if (c == (cns[vpos]>>16&3)) c = 2;
|
|
|
645 else c = 0;
|
|
|
646 // write to seqs
|
|
|
647 key = X31_hash_string(bam1_qname(p->b));
|
|
|
648 k = kh_put(64, seqs, key, &tmp);
|
|
|
649 f = &kh_val(seqs, k);
|
|
|
650 if (tmp == 0) { // present in the hash table
|
|
|
651 if (vpos - f->vpos + 1 < MAX_VARS) {
|
|
|
652 f->vlen = vpos - f->vpos + 1;
|
|
|
653 f->seq[f->vlen-1] = c;
|
|
|
654 f->end = bam_calend(&p->b->core, bam1_cigar(p->b));
|
|
|
655 }
|
|
|
656 dophase = 0;
|
|
|
657 } else { // absent
|
|
|
658 memset(f->seq, 0, MAX_VARS);
|
|
|
659 f->beg = p->b->core.pos;
|
|
|
660 f->end = bam_calend(&p->b->core, bam1_cigar(p->b));
|
|
|
661 f->vpos = vpos, f->vlen = 1, f->seq[0] = c, f->single = f->phased = f->flip = f->ambig = 0;
|
|
|
662 }
|
|
|
663 }
|
|
|
664 if (dophase) {
|
|
|
665 seqs = shrink_hash(seqs);
|
|
|
666 phase(&g, h->target_name[tid], vpos, cns, seqs);
|
|
|
667 update_vpos(vpos, seqs);
|
|
|
668 cns[0] = cns[vpos];
|
|
|
669 vpos = 0;
|
|
|
670 }
|
|
|
671 ++vpos;
|
|
|
672 }
|
|
|
673 if (tid >= 0) phase(&g, h->target_name[tid], vpos, cns, seqs);
|
|
|
674 bam_header_destroy(h);
|
|
|
675 bam_plp_destroy(iter);
|
|
|
676 bam_close(g.fp);
|
|
|
677 kh_destroy(64, seqs);
|
|
|
678 kh_destroy(set64, set);
|
|
|
679 free(cns);
|
|
|
680 errmod_destroy(em);
|
|
|
681 free(bases);
|
|
|
682 if (g.pre) {
|
|
|
683 for (c = 0; c <= 2; ++c) bam_close(g.out[c]);
|
|
|
684 free(g.pre); free(g.b);
|
|
|
685 }
|
|
|
686 return 0;
|
|
|
687 }
|
