X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=bam_plcmd.c;h=e2601ecc3c26b41051d0148a9cf9696036d297fb;hb=a7dfcd368cda94d5847d3f2b42d4288399d77835;hp=98a1e493e2cd824d55d240f588bd2aaa5b067f6f;hpb=1fff3215f3a8beeb98a8025af7a86c7839f0b100;p=samtools.git diff --git a/bam_plcmd.c b/bam_plcmd.c index 98a1e49..e2601ec 100644 --- a/bam_plcmd.c +++ b/bam_plcmd.c @@ -1,17 +1,23 @@ +#include #include #include #include -#include "bam.h" +#include "sam.h" #include "faidx.h" #include "bam_maqcns.h" #include "khash.h" #include "glf.h" -KHASH_SET_INIT_INT64(64) +#include "kstring.h" + +typedef int *indel_list_t; +KHASH_MAP_INIT_INT64(64, indel_list_t) #define BAM_PLF_SIMPLE 0x01 #define BAM_PLF_CNS 0x02 #define BAM_PLF_INDEL_ONLY 0x04 #define BAM_PLF_GLF 0x08 +#define BAM_PLF_VAR_ONLY 0x10 +#define BAM_PLF_2ND 0x20 typedef struct { bam_header_t *h; @@ -23,31 +29,58 @@ typedef struct { int tid, len, last_pos; int mask; char *ref; - glfFile fp; // for glf output only + glfFile fp_glf; // for glf output only } pu_data_t; -char **bam_load_pos(const char *fn, int *_n); +char **__bam_get_lines(const char *fn, int *_n); void bam_init_header_hash(bam_header_t *header); int32_t bam_get_tid(const bam_header_t *header, const char *seq_name); static khash_t(64) *load_pos(const char *fn, bam_header_t *h) { - int n, tmp, i; - char **list, *s; - uint64_t x; + char **list; + int i, j, n, *fields, max_fields; khash_t(64) *hash; bam_init_header_hash(h); - list = bam_load_pos(fn, &n); + list = __bam_get_lines(fn, &n); hash = kh_init(64); + max_fields = 0; fields = 0; for (i = 0; i < n; ++i) { - x = (uint64_t)bam_get_tid(h, list[i]) << 32; - s = list[i]; - while (*s++); - x |= *((uint32_t*)s) - 1; - kh_put(64, hash, x, &tmp); - free(list[i]); + char *str = list[i]; + int chr, n_fields, ret; + khint_t k; + uint64_t x; + n_fields = ksplit_core(str, 0, &max_fields, &fields); + if (n_fields < 2) continue; + chr = bam_get_tid(h, str + fields[0]); + if (chr < 0) { + fprintf(stderr, "[load_pos] unknown reference sequence name: %s\n", str + fields[0]); + continue; + } + x = (uint64_t)chr << 32 | (atoi(str + fields[1]) - 1); + k = kh_put(64, hash, x, &ret); + if (ret == 0) { + fprintf(stderr, "[load_pos] position %s:%s has been loaded.\n", str+fields[0], str+fields[1]); + continue; + } + kh_val(hash, k) = 0; + if (n_fields > 2) { + // count + for (j = 2; j < n_fields; ++j) { + char *s = str + fields[j]; + if ((*s != '+' && *s != '-') || !isdigit(s[1])) break; + } + if (j > 2) { // update kh_val() + int *q, y, z; + q = kh_val(hash, k) = (int*)calloc(j - 1, sizeof(int)); + q[0] = j - 2; z = j; y = 1; + for (j = 2; j < z; ++j) + q[y++] = atoi(str + fields[j]); + } + } + free(str); } - free(list); + free(list); free(fields); return hash; } @@ -56,21 +89,26 @@ static int glt3_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *pu, { pu_data_t *d = (pu_data_t*)data; bam_maqindel_ret_t *r = 0; - int rb; + int rb, *proposed_indels = 0; glf1_t *g; glf3_t *g3; + if (d->fai == 0) { fprintf(stderr, "[glt3_func] reference sequence is required for generating GLT. Abort!\n"); exit(1); } - if (d->hash && kh_get(64, d->hash, (uint64_t)tid<<32|pos) == kh_end(d->hash)) return 0; + if (d->hash) { // only output a list of sites + khint_t k = kh_get(64, d->hash, (uint64_t)tid<<32|pos); + if (k == kh_end(d->hash)) return 0; + proposed_indels = kh_val(d->hash, k); + } g3 = glf3_init1(); if (d->fai && (int)tid != d->tid) { if (d->ref) { // then write the end mark g3->rtype = GLF3_RTYPE_END; - glf3_write1(d->fp, g3); + glf3_write1(d->fp_glf, g3); } - glf3_ref_write(d->fp, d->h->target_name[tid], d->h->target_len[tid]); // write reference + glf3_ref_write(d->fp_glf, d->h->target_name[tid], d->h->target_len[tid]); // write reference free(d->ref); d->ref = fai_fetch(d->fai, d->h->target_name[tid], &d->len); d->tid = tid; @@ -82,8 +120,10 @@ static int glt3_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *pu, g3->rtype = GLF3_RTYPE_SUB; g3->offset = pos - d->last_pos; d->last_pos = pos; - glf3_write1(d->fp, g3); - r = bam_maqindel(n, pos, d->ido, pu, d->ref); + glf3_write1(d->fp_glf, g3); + if (proposed_indels) + r = bam_maqindel(n, pos, d->ido, pu, d->ref, proposed_indels[0], proposed_indels+1); + else r = bam_maqindel(n, pos, d->ido, pu, d->ref, 0, 0); if (r) { // then write indel line int het = 3 * n, min; min = het; @@ -102,7 +142,7 @@ static int glt3_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *pu, g3->max_len = (abs(r->indel1) > abs(r->indel2)? abs(r->indel1) : abs(r->indel2)) + 1; g3->indel_seq[0] = strdup(r->s[0]+1); g3->indel_seq[1] = strdup(r->s[1]+1); - glf3_write1(d->fp, g3); + glf3_write1(d->fp_glf, g3); bam_maqindel_ret_destroy(r); } free(g); @@ -114,45 +154,72 @@ static int pileup_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *p { pu_data_t *d = (pu_data_t*)data; bam_maqindel_ret_t *r = 0; - int i, j, rb, max_mapq = 0; - uint32_t x; - if (d->hash && kh_get(64, d->hash, (uint64_t)tid<<32|pos) == kh_end(d->hash)) return 0; + int i, j, rb, rms_mapq = -1, *proposed_indels = 0; + uint64_t rms_aux; + uint32_t cns = 0; + + // if GLF is required, suppress -c completely if (d->format & BAM_PLF_GLF) return glt3_func(tid, pos, n, pu, data); + // if d->hash is initialized, only output the sites in the hash table + if (d->hash) { + khint_t k = kh_get(64, d->hash, (uint64_t)tid<<32|pos); + if (k == kh_end(d->hash)) return 0; + proposed_indels = kh_val(d->hash, k); + } + // update d->ref if necessary if (d->fai && (int)tid != d->tid) { free(d->ref); d->ref = fai_fetch(d->fai, d->h->target_name[tid], &d->len); d->tid = tid; } rb = (d->ref && (int)pos < d->len)? d->ref[pos] : 'N'; + // when the indel-only mode is asked for, return if no reads mapped with indels if (d->format & BAM_PLF_INDEL_ONLY) { for (i = 0; i < n; ++i) if (pu[i].indel != 0) break; if (i == n) return 0; } + // call the consensus and indel + if (d->format & BAM_PLF_CNS) // call consensus + cns = bam_maqcns_call(n, pu, d->c); + if ((d->format & (BAM_PLF_CNS|BAM_PLF_INDEL_ONLY)) && d->ref && pos < d->len) { // call indels + if (proposed_indels) // the first element gives the size of the array + r = bam_maqindel(n, pos, d->ido, pu, d->ref, proposed_indels[0], proposed_indels+1); + else r = bam_maqindel(n, pos, d->ido, pu, d->ref, 0, 0); + } + // when only variant sites are asked for, test if the site is a variant + if ((d->format & BAM_PLF_CNS) && (d->format & BAM_PLF_VAR_ONLY)) { + if (!(bam_nt16_table[rb] != 15 && cns>>28 != bam_nt16_table[rb])) { // not a SNP + if (!(r && (r->gt == 2 || strcmp(r->s[r->gt], "*")))) { // not an indel + if (r) bam_maqindel_ret_destroy(r); + return 0; + } + } + } + // print the first 3 columns printf("%s\t%d\t%c\t", d->h->target_name[tid], pos + 1, rb); - if (d->format & BAM_PLF_CNS) { // consensus + // print consensus information if required + if (d->format & BAM_PLF_CNS) { int ref_q, rb4 = bam_nt16_table[rb]; - x = bam_maqcns_call(n, pu, d->c); ref_q = 0; - if (rb4 != 15 && x>>28 != 15 && x>>28 != rb4) { // a SNP - ref_q = ((x>>24&0xf) == rb4)? x>>8&0xff : (x>>8&0xff) + (x&0xff); + if (rb4 != 15 && cns>>28 != 15 && cns>>28 != rb4) { // a SNP + ref_q = ((cns>>24&0xf) == rb4)? cns>>8&0xff : (cns>>8&0xff) + (cns&0xff); if (ref_q > 255) ref_q = 255; } - printf("%c\t%d\t%d\t%d\t", bam_nt16_rev_table[x>>28], x>>8&0xff, ref_q, x>>16&0xff); + rms_mapq = cns>>16&0xff; + printf("%c\t%d\t%d\t%d\t", bam_nt16_rev_table[cns>>28], cns>>8&0xff, ref_q, rms_mapq); } - if (d->format & (BAM_PLF_CNS|BAM_PLF_INDEL_ONLY)) { - if (d->ref) // indel calling - r = bam_maqindel(n, pos, d->ido, pu, d->ref); - } - // pileup strings + // print pileup sequences printf("%d\t", n); + rms_aux = 0; // we need to recalculate rms_mapq when -c is not flagged on the command line for (i = 0; i < n; ++i) { const bam_pileup1_t *p = pu + i; - if (max_mapq < p->b->core.qual) max_mapq = p->b->core.qual; + int tmp = p->b->core.qual < d->c->cap_mapQ? p->b->core.qual : d->c->cap_mapQ; + rms_aux += tmp * tmp; if (p->is_head) printf("^%c", p->b->core.qual > 93? 126 : p->b->core.qual + 33); if (!p->is_del) { int c = bam_nt16_rev_table[bam1_seqi(bam1_seq(p->b), p->qpos)]; - if (toupper(c) == toupper(rb)) c = bam1_strand(p->b)? ',' : '.'; + if (c == '=' || toupper(c) == toupper(rb)) c = bam1_strand(p->b)? ',' : '.'; else c = bam1_strand(p->b)? tolower(c) : toupper(c); putchar(c); if (p->indel > 0) { @@ -171,13 +238,33 @@ static int pileup_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *p } else putchar('*'); if (p->is_tail) putchar('$'); } + // finalize rms_mapq + rms_aux = (uint64_t)(sqrt((double)rms_aux / n) + .499); + if (rms_mapq < 0) rms_mapq = rms_aux; putchar('\t'); + // print quality for (i = 0; i < n; ++i) { const bam_pileup1_t *p = pu + i; int c = bam1_qual(p->b)[p->qpos] + 33; if (c > 126) c = 126; putchar(c); } + if (d->format & BAM_PLF_2ND) { // print 2nd calls and qualities + const unsigned char *q; + putchar('\t'); + for (i = 0; i < n; ++i) { + const bam_pileup1_t *p = pu + i; + q = bam_aux_get(p->b, "E2"); + putchar(q? q[p->qpos + 1] : 'N'); + } + putchar('\t'); + for (i = 0; i < n; ++i) { + const bam_pileup1_t *p = pu + i; + q = bam_aux_get(p->b, "U2"); + putchar(q? q[p->qpos + 1] : '!'); + } + } + // print mapping quality if -s is flagged on the command line if (d->format & BAM_PLF_SIMPLE) { putchar('\t'); for (i = 0; i < n; ++i) { @@ -187,15 +274,18 @@ static int pileup_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *p } } putchar('\n'); - if (r) { // then print indel line + // print the indel line if r has been calculated. This only happens if: + // a) -c or -i are flagged, AND b) the reference sequence is available + if (r) { printf("%s\t%d\t*\t", d->h->target_name[tid], pos + 1); if (r->gt < 2) printf("%s/%s\t", r->s[r->gt], r->s[r->gt]); else printf("%s/%s\t", r->s[0], r->s[1]); printf("%d\t%d\t", r->q_cns, r->q_ref); - printf("%d\t%d\t", max_mapq, n); + printf("%d\t%d\t", rms_mapq, n); printf("%s\t%s\t", r->s[0], r->s[1]); //printf("%d\t%d\t", r->gl[0], r->gl[1]); - printf("%d\t%d\t%d\t%d\n", r->cnt1, r->cnt2, r->cnt_ambi, r->cnt_anti); + printf("%d\t%d\t%d\t", r->cnt1, r->cnt2, r->cnt_anti); + printf("%d\t%d\n", r->cnt_ref, r->cnt_ambi); bam_maqindel_ret_destroy(r); } return 0; @@ -203,41 +293,52 @@ static int pileup_func(uint32_t tid, uint32_t pos, int n, const bam_pileup1_t *p int bam_pileup(int argc, char *argv[]) { - int c; + int c, is_SAM = 0; char *fn_list = 0, *fn_fa = 0, *fn_pos = 0; pu_data_t *d = (pu_data_t*)calloc(1, sizeof(pu_data_t)); d->tid = -1; d->mask = BAM_DEF_MASK; d->c = bam_maqcns_init(); d->ido = bam_maqindel_opt_init(); - while ((c = getopt(argc, argv, "st:f:cT:N:r:l:im:gI:G:")) >= 0) { + while ((c = getopt(argc, argv, "st:f:cT:N:r:l:im:gI:G:vM:S2a")) >= 0) { switch (c) { + case 'a': d->c->is_soap = 1; break; case 's': d->format |= BAM_PLF_SIMPLE; break; case 't': fn_list = strdup(optarg); break; case 'l': fn_pos = strdup(optarg); break; case 'f': fn_fa = strdup(optarg); break; case 'T': d->c->theta = atof(optarg); break; case 'N': d->c->n_hap = atoi(optarg); break; - case 'r': d->c->het_rate = atoi(optarg); break; + case 'r': d->c->het_rate = atof(optarg); break; + case 'M': d->c->cap_mapQ = atoi(optarg); break; case 'c': d->format |= BAM_PLF_CNS; break; case 'i': d->format |= BAM_PLF_INDEL_ONLY; break; - case 'm': d->mask = atoi(optarg); break; + case 'v': d->format |= BAM_PLF_VAR_ONLY; break; + case 'm': d->mask = strtol(optarg, 0, 0); break; case 'g': d->format |= BAM_PLF_GLF; break; + case '2': d->format |= BAM_PLF_2ND; break; case 'I': d->ido->q_indel = atoi(optarg); break; case 'G': d->ido->r_indel = atof(optarg); break; + case 'S': is_SAM = 1; break; default: fprintf(stderr, "Unrecognizd option '-%c'.\n", c); return 1; } } + if (fn_list) is_SAM = 1; if (optind == argc) { fprintf(stderr, "\n"); - fprintf(stderr, "Usage: bamtk pileup [options] |\n\n"); + fprintf(stderr, "Usage: samtools pileup [options] |\n\n"); fprintf(stderr, "Option: -s simple (yet incomplete) pileup format\n"); + fprintf(stderr, " -S the input is in SAM\n"); + fprintf(stderr, " -a use the SOAPsnp model for SNP calling\n"); + fprintf(stderr, " -2 output the 2nd best call and quality\n"); fprintf(stderr, " -i only show lines/consensus with indels\n"); fprintf(stderr, " -m INT filtering reads with bits in INT [%d]\n", d->mask); - fprintf(stderr, " -t FILE list of reference sequences (assume the input is in SAM)\n"); + fprintf(stderr, " -M INT cap mapping quality at INT [%d]\n", d->c->cap_mapQ); + fprintf(stderr, " -t FILE list of reference sequences (force -S)\n"); fprintf(stderr, " -l FILE list of sites at which pileup is output\n"); fprintf(stderr, " -f FILE reference sequence in the FASTA format\n\n"); fprintf(stderr, " -c output the maq consensus sequence\n"); - fprintf(stderr, " -g output in the extended GLT format (suppressing -c/-i/-s)\n"); + fprintf(stderr, " -v print variants only (for -c)\n"); + fprintf(stderr, " -g output in the GLFv3 format (suppressing -c/-i/-s)\n"); fprintf(stderr, " -T FLOAT theta in maq consensus calling model (for -c/-g) [%f]\n", d->c->theta); fprintf(stderr, " -N INT number of haplotypes in the sample (for -c/-g) [%d]\n", d->c->n_hap); fprintf(stderr, " -r FLOAT prior of a difference between two haplotypes (for -c/-g) [%f]\n", d->c->het_rate); @@ -248,43 +349,40 @@ int bam_pileup(int argc, char *argv[]) return 1; } if (fn_fa) d->fai = fai_load(fn_fa); - free(fn_fa); - bam_maqcns_prepare(d->c); - if (d->format & BAM_PLF_GLF) { + if (d->format & (BAM_PLF_CNS|BAM_PLF_GLF)) bam_maqcns_prepare(d->c); // consensus calling + if (d->format & BAM_PLF_GLF) { // for glf output glf3_header_t *h; h = glf3_header_init(); - d->fp = bgzf_fdopen(fileno(stdout), "w"); - glf3_header_write(d->fp, h); + d->fp_glf = bgzf_fdopen(fileno(stdout), "w"); + glf3_header_write(d->fp_glf, h); glf3_header_destroy(h); } - if (fn_list) { - tamFile fp; - bam1_t *b; - int ret; - bam_plbuf_t *buf = bam_plbuf_init(pileup_func, d); - bam_plbuf_set_mask(buf, d->mask); - d->h = sam_header_read2(fn_list); - if (fn_pos) d->hash = load_pos(fn_pos, d->h); - fp = sam_open(argv[optind]); - b = (bam1_t*)calloc(1, sizeof(bam1_t)); - while ((ret = sam_read1(fp, d->h, b)) >= 0) - bam_plbuf_push(b, buf); - bam_plbuf_push(0, buf); - bam_plbuf_destroy(buf); - bam_destroy1(b); - sam_close(fp); - } else { - bamFile fp; - fp = (strcmp(argv[optind], "-") == 0)? bam_dopen(fileno(stdin), "r") : bam_open(argv[optind], "r"); - d->h = bam_header_read(fp); + if (d->fai == 0 && (d->format & (BAM_PLF_CNS|BAM_PLF_INDEL_ONLY))) + fprintf(stderr, "[bam_pileup] indels will not be called when -f is absent.\n"); + if (fn_fa && is_SAM && fn_list == 0) fn_list = samfaipath(fn_fa); + + { + samfile_t *fp; + fp = is_SAM? samopen(argv[optind], "r", fn_list) : samopen(argv[optind], "rb", 0); + if (fp == 0 || fp->header == 0) { + fprintf(stderr, "[bam_pileup] fail to read the header: non-exisiting file or wrong format.\n"); + return 1; + } + d->h = fp->header; if (fn_pos) d->hash = load_pos(fn_pos, d->h); - bam_pileup_file(fp, d->mask, pileup_func, d); - bam_close(fp); + sampileup(fp, d->mask, pileup_func, d); + samclose(fp); // d->h will be destroyed here + } + + // free + if (d->format & BAM_PLF_GLF) bgzf_close(d->fp_glf); + if (fn_pos) { // free the hash table + khint_t k; + for (k = kh_begin(d->hash); k < kh_end(d->hash); ++k) + if (kh_exist(d->hash, k)) free(kh_val(d->hash, k)); + kh_destroy(64, d->hash); } - if (d->format & BAM_PLF_GLF) bgzf_close(d->fp); - free(fn_pos); free(fn_list); - kh_destroy(64, d->hash); - bam_header_destroy(d->h); + free(fn_pos); free(fn_list); free(fn_fa); if (d->fai) fai_destroy(d->fai); bam_maqcns_destroy(d->c); free(d->ido); free(d->ref); free(d);