X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=bam_plcmd.c;h=708e63e7e8b600baf166e23979990711d9bf5829;hb=7795f54fa8aed1e0b1edff61890e9fe25fdb7fe9;hp=665304f9019621a8734c0651db57a27a26c49fb2;hpb=961c3fe158b5a9777bc8c061425ca3cb15d8687c;p=samtools.git

diff --git a/bam_plcmd.c b/bam_plcmd.c
index 665304f..708e63e 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"
@@ -343,9 +344,10 @@ int bam_pileup(int argc, char *argv[])
 	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 = 0; break;
+		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;
@@ -380,7 +382,7 @@ 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 MAQ 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);
@@ -389,7 +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);
@@ -448,30 +450,38 @@ int bam_pileup(int argc, char *argv[])
  ***********/
 
 typedef struct {
-	char *reg;
+	int vcf, max_mq, min_mq;
+	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*));
@@ -480,6 +490,7 @@ 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) {
@@ -504,35 +515,102 @@ 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;
+		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);
 	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; // the reference allele frequency
+			int j, _ref, _alt, _ref0, depth, rms_q, _ref0b;
+			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 flast = f;
+				for (j = 0; j < 10; ++j) {
+					f = mc_freq_iter(flast, ma);
+					if (fabs(f - flast) < 1e-3) break;
+					flast = f;
+				}
+			}
+			printf("%s\t%d\t.\t%c\t", h->target_name[tid], pos + 1, _ref0b);
+			if (_ref0 == _ref) putchar("ACGTN"[_alt]);
+			else printf("%c,%c", "ACGTN"[_ref], "ACGTN"[_alt]);
+			printf("\t0\t"); // FIXME: currently these not available
+			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;AF=%.3lg", depth, rms_q, f0<0.?0.:1.-f);
+			printf("\tDP");
+			// output genotype information; FIXME: to be implmented...
+			for (i = 0; i < n; ++i)
+				printf("\t%d", n_plp[i]);
+			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) {
@@ -549,17 +627,22 @@ 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;
+	while ((c = getopt(argc, argv, "f:r:l:VM:q:")) >= 0) {
 		switch (c) {
 		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.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, "Usage: samtools mpileup [-r reg] [-f in.fa] [-l pos] in1.bam [in2.bam [...]]\n");
 		return 1;
 	}
 	mpileup(&mplp, argc - optind, argv + optind);