X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=bam_plcmd.c;h=2c3178c23874c27750bbaa3c3296d2bace171b4a;hb=c0490a669e9f1f9257e4b02a263e92e842bf5357;hp=92023b87f92e4c293c7bb4409cac63373668c017;hpb=aee93c83c1c7ff30c2c959a14877cf8a4e2d8f92;p=samtools.git

diff --git a/bam_plcmd.c b/bam_plcmd.c
index 92023b8..2c3178c 100644
--- a/bam_plcmd.c
+++ b/bam_plcmd.c
@@ -5,6 +5,8 @@
 #include "sam.h"
 #include "faidx.h"
 #include "bam_maqcns.h"
+#include "bam_mcns.h"
+#include "bam2bcf.h"
 #include "khash.h"
 #include "glf.h"
 #include "kstring.h"
@@ -163,18 +165,18 @@ static void pileup_seq(const bam_pileup1_t *p, int pos, int ref_len, const char
 	if (p->is_head) printf("^%c", p->b->core.qual > 93? 126 : p->b->core.qual + 33);
 	if (!p->is_del) {
 		int j, rb, c = bam_nt16_rev_table[bam1_seqi(bam1_seq(p->b), p->qpos)];
-		rb = (ref && (int)pos < ref_len)? ref[pos] : 'N';
+		rb = (ref && pos < ref_len)? ref[pos] : 'N';
 		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) {
-			putchar('+'); putw(p->indel, stdout);
+			printf("+%d", p->indel);
 			for (j = 1; j <= p->indel; ++j) {
 				c = bam_nt16_rev_table[bam1_seqi(bam1_seq(p->b), p->qpos + j)];
 				putchar(bam1_strand(p->b)? tolower(c) : toupper(c));
 			}
 		} else if (p->indel < 0) {
-			putw(p->indel, stdout);
+			printf("%d", p->indel);
 			for (j = 1; j <= -p->indel; ++j) {
 				c = (ref && (int)pos+j < ref_len)? ref[pos+j] : 'N';
 				putchar(bam1_strand(p->b)? tolower(c) : toupper(c));
@@ -341,17 +343,19 @@ 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 +383,7 @@ int bam_pileup(int argc, char *argv[])
 		fprintf(stderr, "Usage:  samtools pileup [options] <in.bam>|<in.sam>\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 +392,7 @@ 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);
@@ -446,60 +450,324 @@ int bam_pileup(int argc, char *argv[])
  * mpileup *
  ***********/
 
-static int mpileup(int n, char **fn)
+#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 max_mq, min_mq, prior_type, flag, min_baseQ;
+	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;
+	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)
 {
-	bamFile *fp;
-	int i, tid, pos, *n_plp;
+	mplp_aux_t **data;
+	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;
-	fp = calloc(n, sizeof(void*));
+	uint64_t N = 0;
+	char *ref;
+	khash_t(64) *hash = 0;
+
+	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;
-		fp[i] = bam_open(fn[i], "r");
-		h_tmp = bam_header_read(fp[i]);
+		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) {
+			int beg, end;
+			bam_index_t *idx;
+			idx = bam_index_load(fn[i]);
+			if (idx == 0) {
+				fprintf(stderr, "[%s] fail to load index for %d-th input.\n", __func__, i+1);
+				exit(1);
+			}
+			if (bam_parse_region(h_tmp, conf->reg, &tid, &beg, &end) < 0) {
+				fprintf(stderr, "[%s] malformatted region or wrong seqname for %d-th input.\n", __func__, i+1);
+				exit(1);
+			}
+			if (i == 0) beg0 = beg, end0 = end;
+			data[i]->iter = bam_iter_query(idx, tid, beg, end);
+			bam_index_destroy(idx);
+		}
 		if (i == 0) h = h_tmp;
 		else {
 			// FIXME: to check consistency
 			bam_header_destroy(h_tmp);
 		}
 	}
-	iter = bam_mplp_init(n, bam_read1, fp);
-	while (bam_mplp_auto(iter, &tid, &pos, n_plp, plp) > 0) {
-		printf("%s\t%d\tN", h->target_name[tid], pos + 1);
+	if (conf->fn_pos) hash = load_pos(conf->fn_pos, h);
+	// write the VCF header
+	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) {
-			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, 0, 0);
-				putchar('\t');
+			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=<ID=DP,Number=1,Type=Integer,Description=\"Total read depth\">");
+		puts("##INFO=<ID=AF,Number=1,Type=Float,Description=\"Non-reference allele frequency \\argmax_f P(D|f)\">");
+		puts("##INFO=<ID=AFE,Number=1,Type=Float,Description=\"Expected non-reference allele frequency\">");
+		puts("##FILTER=<ID=Q13,Description=\"All min{baseQ,mapQ} below 13\">");
+		puts("##FILTER=<ID=FPE,Description=\"Floating point error\">");
+		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->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) {
+		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;
+		}
+		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]];
+			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->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) {
+				putchar("ACGTN"[r.alt]);
+				if (r.alt2 >= 0 && r.alt2 < 4) printf(",%c", "ACGT"[r.alt2]);
+			} else putchar('.');
+			printf("\t%d\t", qref);
+			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];
 				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 (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) {
+				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');
 	}
+	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);
-	for (i = 0; i < n; ++i) bam_close(fp[i]);
-	free(fp); free(plp);
+	for (i = 0; i < n; ++i) {
+		bam_close(data[i]->fp);
+		if (data[i]->iter) bam_iter_destroy(data[i]->iter);
+		free(data[i]);
+	}
+	free(data); free(plp); free(ref); free(n_plp);
 	return 0;
 }
 
 int bam_mpileup(int argc, char *argv[])
 {
+	int c;
+	mplp_conf_t mplp;
+	memset(&mplp, 0, sizeof(mplp_conf_t));
+	mplp.max_mq = 60;
+	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 '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 <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(argc - 1, argv + 1);
+	mpileup(&mplp, argc - optind, argv + optind);
+	free(mplp.reg);
+	if (mplp.fai) fai_destroy(mplp.fai);
 	return 0;
 }