14 /* Note that although the two matrics have 10 parameters in total, only 4
15 * (probably 3) are free. Changing the scoring matrices in a sort of symmetric
16 * way will not change the result. */
17 static score_param_t g_param = { {{0,0,0},{-4,1,6}}, {{0,-14000}, {0,0}} };
20 int min_baseQ, tid, max_bases;
29 static uint16_t gencns(ct_t *g, int n, const bam_pileup1_t *plp)
31 int i, j, ret, tmp, k, sum[4], qual;
33 if (n > g->max_bases) { // enlarge g->bases
35 kroundup32(g->max_bases);
36 g->bases = realloc(g->bases, g->max_bases * 2);
38 for (i = k = 0; i < n; ++i) {
39 const bam_pileup1_t *p = plp + i;
42 if (p->is_refskip || p->is_del) continue;
43 baseQ = bam1_qual(p->b)[p->qpos];
44 if (baseQ < g->min_baseQ) continue;
46 b = bam_nt16_nt4_table[bam1_seqi(seq, p->qpos)];
48 q = baseQ < p->b->core.qual? baseQ : p->b->core.qual;
49 g->bases[k++] = q<<5 | bam1_strand(p->b)<<4 | b;
52 errmod_cal(g->em, k, 4, g->bases, q);
53 for (i = 0; i < 4; ++i) sum[i] = (int)(q[i<<2|i] + .499) << 2 | i;
54 for (i = 1; i < 4; ++i) // insertion sort
55 for (j = i; j > 0 && sum[j] < sum[j-1]; --j)
56 tmp = sum[j], sum[j] = sum[j-1], sum[j-1] = tmp;
57 qual = (sum[1]>>2) - (sum[0]>>2);
59 ret = (qual < 63? qual : 63) << 2 | (sum[0]&3);
63 static void process_cns(bam_header_t *h, int tid, int l, uint16_t *cns)
65 int i, f[2][2], *prev, *curr, *swap_tmp, s;
66 uint8_t *b; // backtrack array
68 f[0][0] = f[0][1] = 0;
69 prev = f[0]; curr = f[1];
70 // fill the backtrack matrix
71 for (i = 0; i < l; ++i) {
72 int c = (cns[i] == 0)? 0 : (cns[i]>>8 == 0)? 1 : 2;
75 tmp0 = prev[0] + g_param.e[0][c] + g_param.p[0][0]; // (s[i+1],s[i])=(0,0)
76 tmp1 = prev[1] + g_param.e[0][c] + g_param.p[1][0]; // (0,1)
77 if (tmp0 > tmp1) curr[0] = tmp0, b[i] = 0;
78 else curr[0] = tmp1, b[i] = 1;
80 tmp0 = prev[0] + g_param.e[1][c] + g_param.p[0][1]; // (s[i+1],s[i])=(1,0)
81 tmp1 = prev[1] + g_param.e[1][c] + g_param.p[1][1]; // (1,1)
82 if (tmp0 > tmp1) curr[1] = tmp0, b[i] |= 0<<1;
83 else curr[1] = tmp1, b[i] |= 1<<1;
85 swap_tmp = prev; prev = curr; curr = swap_tmp;
88 s = prev[0] > prev[1]? 0 : 1;
89 for (i = l - 1; i > 0; --i) {
94 for (i = 0, s = -1; i <= l; ++i) {
95 if (i == l || ((b[i]>>2&3) == 0 && s >= 0)) {
98 printf("%s:%d-%d\t0\t%s\t%d\t60\t%dM\t*\t0\t0\t", h->target_name[tid], s+1, i, h->target_name[tid], s+1, i-s);
99 for (j = s; j < i; ++j) {
101 if (c == 0) putchar('N');
102 else putchar("ACGT"[c&3]);
105 for (j = s; j < i; ++j)
106 putchar(33 + (cns[j]>>8>>2));
109 //if (s >= 0) printf("%s\t%d\t%d\t%d\n", h->target_name[tid], s, i, i - s);
111 } else if ((b[i]>>2&3) && s < 0) s = i;
116 static int read_aln(void *data, bam1_t *b)
118 extern int bam_prob_realn_core(bam1_t *b, const char *ref, int flag);
119 ct_t *g = (ct_t*)data;
121 ret = bam_read1(g->fp, b);
122 if (ret >= 0 && g->fai && b->core.tid >= 0 && (b->core.flag&4) == 0) {
123 if (b->core.tid != g->tid) { // then load the sequence
125 g->ref = fai_fetch(g->fai, g->h->target_name[b->core.tid], &len);
126 g->tid = b->core.tid;
128 bam_prob_realn_core(b, g->ref, 1<<1|1);
133 int main_cut_target(int argc, char *argv[])
135 int c, tid, pos, n, lasttid = -1, lastpos = -1, l, max_l;
136 const bam_pileup1_t *p;
141 memset(&g, 0, sizeof(ct_t));
142 g.min_baseQ = 13; g.tid = -1;
143 while ((c = getopt(argc, argv, "f:Q:i:o:0:1:2:")) >= 0) {
145 case 'Q': g.min_baseQ = atoi(optarg); break; // quality cutoff
146 case 'i': g_param.p[0][1] = -atoi(optarg); break; // 0->1 transition (in) PENALTY
147 case '0': g_param.e[1][0] = atoi(optarg); break; // emission SCORE
148 case '1': g_param.e[1][1] = atoi(optarg); break;
149 case '2': g_param.e[1][2] = atoi(optarg); break;
150 case 'f': g.fai = fai_load(optarg);
151 if (g.fai == 0) fprintf(stderr, "[%s] fail to load the fasta index.\n", __func__);
155 if (argc == optind) {
156 fprintf(stderr, "Usage: samtools targetcut [-Q minQ] [-i inPen] [-0 em0] [-1 em1] [-2 em2] [-f ref] <in.bam>\n");
159 l = max_l = 0; cns = 0;
160 g.fp = strcmp(argv[optind], "-")? bam_open(argv[optind], "r") : bam_dopen(fileno(stdin), "r");
161 g.h = bam_header_read(g.fp);
162 g.em = errmod_init(1 - ERR_DEP);
163 plp = bam_plp_init(read_aln, &g);
164 while ((p = bam_plp_auto(plp, &tid, &pos, &n)) != 0) {
166 if (tid != lasttid) { // change of chromosome
167 if (cns) process_cns(g.h, lasttid, l, cns);
168 if (max_l < g.h->target_len[tid]) {
169 max_l = g.h->target_len[tid];
171 cns = realloc(cns, max_l * 2);
173 l = g.h->target_len[tid];
174 memset(cns, 0, max_l * 2);
177 cns[pos] = gencns(&g, n, p);
180 process_cns(g.h, lasttid, l, cns);
182 bam_header_destroy(g.h);
183 bam_plp_destroy(plp);
186 fai_destroy(g.fai); free(g.ref);
188 errmod_destroy(g.em);