X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=bam_plcmd.c;h=2c3178c23874c27750bbaa3c3296d2bace171b4a;hb=c0490a669e9f1f9257e4b02a263e92e842bf5357;hp=285477050d19f489d7be34722846eca8e6407e14;hpb=d73a1b3b94255044787c0e4bfeba37df66e99770;p=samtools.git diff --git a/bam_plcmd.c b/bam_plcmd.c index 2854770..2c3178c 100644 --- a/bam_plcmd.c +++ b/bam_plcmd.c @@ -6,6 +6,7 @@ #include "faidx.h" #include "bam_maqcns.h" #include "bam_mcns.h" +#include "bam2bcf.h" #include "khash.h" #include "glf.h" #include "kstring.h" @@ -449,8 +450,17 @@ int bam_pileup(int argc, char *argv[]) * mpileup * ***********/ +#define MPLP_VCF 0x1 +#define MPLP_VAR 0x2 +#define MPLP_AFALL 0x8 +#define MPLP_GLF 0x10 +#define MPLP_NO_COMP 0x20 + +#define MPLP_AFS_BLOCK 0x10000 + typedef struct { - int vcf, max_mq, min_mq, var_only; + int max_mq, min_mq, prior_type, flag, min_baseQ; + double theta; char *reg, *fn_pos; faidx_t *fai; kh_64_t *hash; @@ -475,17 +485,26 @@ static int mplp_func(void *data, bam1_t *b) 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; + uint64_t N = 0; char *ref; khash_t(64) *hash = 0; - // allocate + + mc_aux_t *ma = 0; + + bcf_callaux_t *bca = 0; + bcf_callret1_t *bcr = 0; + bcf_call_t bc; + bcf_t *bp = 0; + + memset(&bc, 0, sizeof(bcf_call_t)); data = calloc(n, sizeof(void*)); plp = calloc(n, sizeof(void*)); n_plp = calloc(n, sizeof(int*)); + // read the header and initialize data for (i = 0; i < n; ++i) { bam_header_t *h_tmp; @@ -517,10 +536,41 @@ static int mpileup(mplp_conf_t *conf, int n, char **fn) } if (conf->fn_pos) hash = load_pos(conf->fn_pos, h); // write the VCF header - if (conf->vcf) { + if (conf->flag & MPLP_GLF) { + kstring_t s; + s.l = s.m = 0; s.s = 0; + bp = bcf_open("-", (conf->flag&MPLP_NO_COMP)? "wu" : "w"); + for (i = 0; i < h->n_targets; ++i) { + kputs(h->target_name[i], &s); + kputc('\0', &s); + } + bp->h.l_nm = s.l; + bp->h.name = malloc(s.l); + memcpy(bp->h.name, s.s, s.l); + s.l = 0; + for (i = 0; i < n; ++i) { + const char *p; + if ((p = strstr(fn[i], ".bam")) != 0) + kputsn(fn[i], p - fn[i], &s); + else kputs(fn[i], &s); + kputc('\0', &s); + } + bp->h.l_smpl = s.l; + bp->h.sname = malloc(s.l); + memcpy(bp->h.sname, s.s, s.l); + bp->h.l_txt = 0; + free(s.s); + bcf_hdr_sync(&bp->h); + bcf_hdr_write(bp); + } else if (conf->flag & MPLP_VCF) { kstring_t s; s.l = s.m = 0; s.s = 0; puts("##fileformat=VCFv4.0"); + puts("##INFO="); + puts("##INFO="); + puts("##INFO="); + puts("##FILTER="); + puts("##FILTER="); kputs("#CHROM\tPOS\tID\tREF\tALT\tQUAL\tFILTER\tINFO\tFORMAT", &s); for (i = 0; i < n; ++i) { const char *p; @@ -533,7 +583,13 @@ static int mpileup(mplp_conf_t *conf, int n, char **fn) free(s.s); } // mpileup - if (conf->vcf) ma = mc_init(n); + if (conf->flag & MPLP_GLF) { + bca = bcf_call_init(-1., conf->min_baseQ); + bcr = calloc(n, sizeof(bcf_callret1_t)); + } else if (conf->flag & MPLP_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) { @@ -548,35 +604,43 @@ static int mpileup(mplp_conf_t *conf, int n, char **fn) if (conf->fai) ref = fai_fetch(conf->fai, h->target_name[tid], &ref_len); ref_tid = tid; } - if (conf->vcf) { - double f0, f, pref = -1.0; // the reference allele frequency - int j, _ref, _alt, _ref0, depth, rms_q, _ref0b, is_var = 0, qref = 0; + if (conf->flag & MPLP_GLF) { + int _ref0, ref16; + bcf1_t *b = calloc(1, sizeof(bcf1_t)); + _ref0 = (ref && pos < ref_len)? ref[pos] : 'N'; + ref16 = bam_nt16_table[_ref0]; + for (i = 0; i < n; ++i) + bcf_call_glfgen(n_plp[i], plp[i], ref16, bca, bcr + i); + bcf_call_combine(n, bcr, ref16, &bc); + bcf_call2bcf(tid, pos, &bc, b); + bcf_write(bp, b); + //fprintf(stderr, "%d,%d,%d\n", b->tid, b->pos, b->l_str); + bcf_destroy(b); + } else if (conf->flag & MPLP_VCF) { + mc_rst_t r; + int j, _ref0, depth, rms_q, _ref0b, is_var = 0, qref = 0, level = 2, tot; uint64_t sqr_sum; _ref0 = _ref0b = (ref && pos < ref_len)? ref[pos] : 'N'; _ref0 = bam_nt16_nt4_table[bam_nt16_table[_ref0]]; - f = f0 = 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); - if (fabs(f - flast) < 1e-3) break; - flast = f; - } - pref = mc_ref_prob(ma); - is_var = (pref < .5); - q = is_var? pref : 1. - pref; + tot = mc_cal(_ref0, n_plp, plp, ma, &r, level); + if (tot) { // has good bases + double q; + is_var = (r.p_ref < .5); + q = is_var? r.p_ref : 1. - r.p_ref; 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; + if ((conf->flag & MPLP_VAR) && !is_var) continue; + ++N; // number of processed lines 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]); + putchar("ACGTN"[r.alt]); + if (r.alt2 >= 0 && r.alt2 < 4) printf(",%c", "ACGT"[r.alt2]); } else putchar('.'); printf("\t%d\t", qref); - if (f0 < 0.) printf("Q13\t"); + if (!tot) printf("Q13\t"); + else if (r.f_exp < 0.) printf("FPE\t"); else printf(".\t"); for (i = depth = 0, sqr_sum = 0; i < n; ++i) { depth += n_plp[i]; @@ -587,14 +651,22 @@ static int mpileup(mplp_conf_t *conf, int n, char **fn) } } rms_q = (int)(sqrt((double)sqr_sum / depth) + .499); - printf("DP=%d;MQ=%d;AF=%.3lg", depth, rms_q, f0<0.?0.:1.-f); - printf("\tGT:GQ:DP"); - // output genotype information; FIXME: to be implmented... - for (i = 0; i < n; ++i) { - int x = mc_call_gt(ma, f, i); - printf("\t%c/%c:%d:%d", "10"[((x&3)==2)], "10"[((x&3)>0)], x>>2, n_plp[i]); + printf("DP=%d;MQ=%d", depth, rms_q); + if (tot) { + printf(";AF=%.3lf", 1. - r.f_em); + if (level >= 2) printf(";AFE=%.3lf", 1-r.f_exp); + if (conf->flag & MPLP_AFALL) + printf(";AF0=%.3lf;AFN=%.3lf", 1-r.f_naive, 1-r.f_nielsen); } + printf("\tGT:GQ:DP"); + if (tot) { + for (i = 0; i < n; ++i) { + int x = mc_call_gt(ma, r.f_exp, 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'); + if (N % MPLP_AFS_BLOCK == 0) mc_dump_afs(ma); } 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) { @@ -616,12 +688,15 @@ static int mpileup(mplp_conf_t *conf, int n, char **fn) putchar('\n'); } } + bcf_close(bp); + if (conf->flag&MPLP_VCF) mc_dump_afs(ma); 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); } + bcf_call_destroy(bca); free(bc.PL); free(bcr); mc_destroy(ma); bam_mplp_destroy(iter); bam_header_destroy(h); @@ -640,22 +715,55 @@ int bam_mpileup(int argc, char *argv[]) mplp_conf_t mplp; memset(&mplp, 0, sizeof(mplp_conf_t)); mplp.max_mq = 60; - while ((c = getopt(argc, argv, "f:r:l:VvM:q:")) >= 0) { + mplp.prior_type = MC_PTYPE_FULL; + mplp.theta = 1e-3; + mplp.min_baseQ = 13; + while ((c = getopt(argc, argv, "gvVcFSP:f:r:l:VM:q:t:Q:u")) >= 0) { switch (c) { + case 't': mplp.theta = atof(optarg); break; + case 'P': + if (strcmp(optarg, "full") == 0) mplp.prior_type = MC_PTYPE_FULL; + else if (strcmp(optarg, "cond2") == 0) mplp.prior_type = MC_PTYPE_COND2; + else if (strcmp(optarg, "flat") == 0) mplp.prior_type = MC_PTYPE_FLAT; + else { + fprintf(stderr, "[%s] unrecognized prior type.\n", __func__); + return 1; + } + break; case 'f': mplp.fai = fai_load(optarg); if (mplp.fai == 0) return 1; break; 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 'g': mplp.flag |= MPLP_GLF; break; + case 'V': + case 'c': mplp.flag |= MPLP_VCF; break; + case 'F': mplp.flag |= MPLP_AFALL; break; + case 'v': mplp.flag |= MPLP_VAR; break; + case 'u': mplp.flag |= MPLP_NO_COMP; break; case 'M': mplp.max_mq = atoi(optarg); break; case 'q': mplp.min_mq = atoi(optarg); break; + case 'Q': mplp.min_baseQ = atoi(optarg); break; } } + if (mplp.flag&MPLP_GLF) mplp.flag &= ~MPLP_VCF; if (argc == 1) { - fprintf(stderr, "Usage: samtools mpileup [-r reg] [-f in.fa] [-l pos] 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, " -P STR prior: full, flat, cond2 [full]\n"); + fprintf(stderr, " -Q INT min base quality [%d]\n", mplp.min_baseQ); + fprintf(stderr, " -c generate VCF output (consensus calling)\n"); + fprintf(stderr, " -g generate GLF output\n"); + fprintf(stderr, " -v show variant sites only\n"); + fprintf(stderr, "\n"); + fprintf(stderr, "Notes: Assuming error independency and diploid individuals.\n\n"); return 1; } mpileup(&mplp, argc - optind, argv + optind);