X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=bam_plcmd.c;h=82fcafc7f9c8836d839aef7ecaf26291ed1f89a0;hb=71355852e510ac7937d552117557b9dbf07cd557;hp=d1ea03f896e9ad433c95d33ba1dfe88ad2354aae;hpb=a958954399757774ee26bfcf0b5b95e9ec9b62f4;p=samtools.git diff --git a/bam_plcmd.c b/bam_plcmd.c index d1ea03f..82fcafc 100644 --- a/bam_plcmd.c +++ b/bam_plcmd.c @@ -5,6 +5,7 @@ #include "sam.h" #include "faidx.h" #include "bam_maqcns.h" +#include "bam_mcns.h" #include "khash.h" #include "glf.h" #include "kstring.h" @@ -341,17 +342,19 @@ int bam_pileup(int argc, char *argv[]) d->max_depth = 0; d->tid = -1; d->mask = BAM_DEF_MASK; d->c = bam_maqcns_init(); + d->c->is_soap = 1; // change the default model d->ido = bam_maqindel_opt_init(); - while ((c = getopt(argc, argv, "st:f:cT:N:r:l:d:im:gI:G:vM:S2aR:P")) >= 0) { + while ((c = getopt(argc, argv, "st:f:cT:N:r:l:d:im:gI:G:vM:S2aR:PA")) >= 0) { switch (c) { case 'a': d->c->is_soap = 1; break; + case 'A': d->c->is_soap = 0; 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 = atof(optarg); break; + case 'r': d->c->het_rate = atof(optarg); d->ido->r_snp = d->c->het_rate; break; case 'M': d->c->cap_mapQ = atoi(optarg); break; case 'd': d->max_depth = atoi(optarg); break; case 'c': d->format |= BAM_PLF_CNS; break; @@ -379,7 +382,7 @@ int bam_pileup(int argc, char *argv[]) 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, " -A use the MAQ 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); @@ -388,7 +391,7 @@ int bam_pileup(int argc, char *argv[]) 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, " -c output the SOAPsnp consensus sequence\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); @@ -447,30 +450,39 @@ int bam_pileup(int argc, char *argv[]) ***********/ typedef struct { - char *reg; + int vcf, max_mq, min_mq, var_only, prior_type; + double theta; + char *reg, *fn_pos; faidx_t *fai; + kh_64_t *hash; } mplp_conf_t; typedef struct { bamFile fp; bam_iter_t iter; + int min_mq; } mplp_aux_t; static int mplp_func(void *data, bam1_t *b) { mplp_aux_t *ma = (mplp_aux_t*)data; - if (ma->iter) return bam_iter_read(ma->fp, ma->iter, b); - return bam_read1(ma->fp, b); + int ret; + do { + ret = ma->iter? bam_iter_read(ma->fp, ma->iter, b) : bam_read1(ma->fp, b); + } while (b->core.qual < ma->min_mq && ret >= 0); + return ret; } static int mpileup(mplp_conf_t *conf, int n, char **fn) { mplp_aux_t **data; + mc_aux_t *ma = 0; int i, tid, pos, *n_plp, beg0 = 0, end0 = 1u<<29, ref_len, ref_tid; const bam_pileup1_t **plp; bam_mplp_t iter; bam_header_t *h = 0; char *ref; + khash_t(64) *hash = 0; // allocate data = calloc(n, sizeof(void*)); plp = calloc(n, sizeof(void*)); @@ -479,6 +491,7 @@ static int mpileup(mplp_conf_t *conf, int n, char **fn) for (i = 0; i < n; ++i) { bam_header_t *h_tmp; data[i] = calloc(1, sizeof(mplp_aux_t)); + data[i]->min_mq = conf->min_mq; data[i]->fp = bam_open(fn[i], "r"); h_tmp = bam_header_read(data[i]->fp); if (conf->reg) { @@ -503,35 +516,127 @@ static int mpileup(mplp_conf_t *conf, int n, char **fn) bam_header_destroy(h_tmp); } } + if (conf->fn_pos) hash = load_pos(conf->fn_pos, h); + // write the VCF header + if (conf->vcf) { + kstring_t s; + s.l = s.m = 0; s.s = 0; + puts("##fileformat=VCFv4.0"); + puts("##INFO="); + puts("##INFO="); + puts("##INFO="); + puts("##FILTER="); + kputs("#CHROM\tPOS\tID\tREF\tALT\tQUAL\tFILTER\tINFO\tFORMAT", &s); + for (i = 0; i < n; ++i) { + const char *p; + kputc('\t', &s); + if ((p = strstr(fn[i], ".bam")) != 0) + kputsn(fn[i], p - fn[i], &s); + else kputs(fn[i], &s); + } + puts(s.s); + free(s.s); + } // mpileup + if (conf->vcf) { + ma = mc_init(n); + mc_init_prior(ma, conf->prior_type, conf->theta); + } ref_tid = -1; ref = 0; iter = bam_mplp_init(n, mplp_func, (void**)data); while (bam_mplp_auto(iter, &tid, &pos, n_plp, plp) > 0) { if (conf->reg && (pos < beg0 || pos >= end0)) continue; // out of the region requested + if (hash) { + khint_t k; + k = kh_get(64, hash, (uint64_t)tid<<32 | pos); + if (k == kh_end(hash)) continue; + } if (tid != ref_tid) { free(ref); if (conf->fai) ref = fai_fetch(conf->fai, h->target_name[tid], &ref_len); ref_tid = tid; } - printf("%s\t%d\t%c", h->target_name[tid], pos + 1, (ref && pos < ref_len)? ref[pos] : 'N'); - for (i = 0; i < n; ++i) { - int j; - printf("\t%d\t", n_plp[i]); - if (n_plp[i] == 0) printf("*\t*"); - else { - for (j = 0; j < n_plp[i]; ++j) - pileup_seq(plp[i] + j, pos, ref_len, ref); - putchar('\t'); + if (conf->vcf) { + double f0, f, fpost, pref = -1.0; // the reference allele frequency + int j, _ref, _alt, _ref0, depth, rms_q, _ref0b, is_var = 0, qref = 0; + uint64_t sqr_sum; + _ref0 = _ref0b = (ref && pos < ref_len)? ref[pos] : 'N'; + _ref0 = bam_nt16_nt4_table[bam_nt16_table[_ref0]]; + f = f0 = fpost = mc_freq0(_ref0, n_plp, plp, ma, &_ref, &_alt); + if (f >= 0.0) { + double q, flast = f; + for (j = 0; j < 10; ++j) { + f = mc_freq_iter(flast, ma); +// printf("%lg->%lg\n", flast, f); + if (fabs(f - flast) < 1e-3) break; + flast = f; + } + pref = mc_ref_prob(ma); + fpost = mc_freq_post(ma); + if (1. - f < 1e-4) f = 1.; + if (1. - fpost < 1e-4) fpost = 1.; + is_var = (pref < .5); + q = is_var? pref : 1. - pref; + if (q < 1e-308) q = 1e-308; + qref = (int)(-3.434 * log(q) + .499); + if (qref > 99) qref = 99; + } + if (conf->var_only && !is_var) continue; + printf("%s\t%d\t.\t%c\t", h->target_name[tid], pos + 1, _ref0b); + if (is_var) { + if (_ref0 == _ref) putchar("ACGTN"[_alt]); + else printf("%c,%c", "ACGTN"[_ref], "ACGTN"[_alt]); + } else putchar('.'); + printf("\t%d\t", qref); + if (f0 < 0.) printf("Q13\t"); + else printf(".\t"); + for (i = depth = 0, sqr_sum = 0; i < n; ++i) { + depth += n_plp[i]; for (j = 0; j < n_plp[i]; ++j) { - const bam_pileup1_t *p = plp[i] + j; - int c = bam1_qual(p->b)[p->qpos] + 33; - if (c > 126) c = 126; - putchar(c); + int q = plp[i][j].b->core.qual; + if (q > conf->max_mq) q = conf->max_mq; + sqr_sum += q * q; } } + rms_q = (int)(sqrt((double)sqr_sum / depth) + .499); + printf("DP=%d;MQ=%d", depth, rms_q); + if (fpost >= 0. && fpost <= 1.) printf(";AF=%.3lg", 1. - fpost); + if (f >= 0. && f <= 1.) printf(";AFEM=%.3lg", 1. - f); + if (fpost >= 0. && fpost <= 1.) { + for (i = 0; i < n; ++i) { + int x = mc_call_gt(ma, fpost, i); + printf("\t%c/%c:%d:%d", "10"[((x&3)==2)], "10"[((x&3)>0)], x>>2, n_plp[i]); + } + } else for (i = 0; i < n; ++i) printf("\t./.:0:0"); + putchar('\n'); + } else { + printf("%s\t%d\t%c", h->target_name[tid], pos + 1, (ref && pos < ref_len)? ref[pos] : 'N'); + for (i = 0; i < n; ++i) { + int j; + printf("\t%d\t", n_plp[i]); + if (n_plp[i] == 0) printf("*\t*"); + else { + for (j = 0; j < n_plp[i]; ++j) + pileup_seq(plp[i] + j, pos, ref_len, ref); + putchar('\t'); + for (j = 0; j < n_plp[i]; ++j) { + const bam_pileup1_t *p = plp[i] + j; + int c = bam1_qual(p->b)[p->qpos] + 33; + if (c > 126) c = 126; + putchar(c); + } + } + } + putchar('\n'); } - putchar('\n'); } + if (hash) { // free the hash table + khint_t k; + for (k = kh_begin(hash); k < kh_end(hash); ++k) + if (kh_exist(hash, k)) free(kh_val(hash, k)); + kh_destroy(64, hash); + } + mc_destroy(ma); bam_mplp_destroy(iter); bam_header_destroy(h); for (i = 0; i < n; ++i) { @@ -548,17 +653,39 @@ int bam_mpileup(int argc, char *argv[]) int c; mplp_conf_t mplp; memset(&mplp, 0, sizeof(mplp_conf_t)); - while ((c = getopt(argc, argv, "f:r:")) >= 0) { + mplp.max_mq = 60; + mplp.prior_type = MC_PTYPE_FULL; + mplp.theta = 1e-3; + while ((c = getopt(argc, argv, "f:r:l:VvM:q:t:2")) >= 0) { switch (c) { + case 't': mplp.theta = atof(optarg); break; + case '2': mplp.prior_type = MC_PTYPE_COND2; break; case 'f': mplp.fai = fai_load(optarg); if (mplp.fai == 0) return 1; break; - case 'r': mplp.reg = strdup(optarg); + case 'r': mplp.reg = strdup(optarg); break; + case 'l': mplp.fn_pos = strdup(optarg); break; + case 'v': mplp.var_only = 1; break; + case 'V': mplp.vcf = 1; break; + case 'M': mplp.max_mq = atoi(optarg); break; + case 'q': mplp.min_mq = atoi(optarg); break; } } if (argc == 1) { - fprintf(stderr, "Usage: samtools mpileup [-r reg] [-f in.fa] in1.bam [in2.bam [...]]\n"); + fprintf(stderr, "\n"); + fprintf(stderr, "Usage: samtools mpileup [options] in1.bam [in2.bam [...]]\n\n"); + fprintf(stderr, "Options: -f FILE reference sequence file [null]\n"); + fprintf(stderr, " -r STR region in which pileup is generated [null]\n"); + fprintf(stderr, " -l FILE list of positions (format: chr pos) [null]\n"); + fprintf(stderr, " -M INT cap mapping quality at INT [%d]\n", mplp.max_mq); + fprintf(stderr, " -q INT filter out alignment with MQ smaller than INT [%d]\n", mplp.min_mq); + fprintf(stderr, " -t FLOAT scaled mutation rate [%lg]\n", mplp.theta); + fprintf(stderr, " -V generate VCF output (SNP calling)\n"); + fprintf(stderr, " -v show variant sites only\n"); + fprintf(stderr, " -2 conditional prior\n"); + fprintf(stderr, "\n"); + fprintf(stderr, "Notes: Assuming error independency and diploid individuals.\n\n"); return 1; } mpileup(&mplp, argc - optind, argv + optind);