X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=bam_plcmd.c;h=cdbf67d861cf5e90418422057901dfab0bf0ee6b;hb=221f82f662dd770a17d1bd2181de46b8c3b3fdfd;hp=80799e3b28af95208796424c31a4bf5b7dd24ba0;hpb=63af098c78ef4163356c281e87bfb467adc8824d;p=samtools.git

diff --git a/bam_plcmd.c b/bam_plcmd.c
index 80799e3..cdbf67d 100644
--- a/bam_plcmd.c
+++ b/bam_plcmd.c
@@ -5,7 +5,6 @@
 #include "sam.h"
 #include "faidx.h"
 #include "bam_maqcns.h"
-#include "bam2bcf.h"
 #include "khash.h"
 #include "glf.h"
 #include "kstring.h"
@@ -376,6 +375,8 @@ int bam_pileup(int argc, char *argv[])
 		default: fprintf(stderr, "Unrecognizd option '-%c'.\n", c); return 1;
 		}
 	}
+	if (d->c->errmod != BAM_ERRMOD_MAQ2) d->c->theta += 0.02;
+	if (d->c->theta > 1.0) d->c->theta = 1.0;
 	if (fn_list) is_SAM = 1;
 	if (optind == argc) {
 		fprintf(stderr, "\n");
@@ -449,8 +450,15 @@ int bam_pileup(int argc, char *argv[])
  * mpileup *
  ***********/
 
+#include <assert.h>
+#include "bam2bcf.h"
+#include "sample.h"
+
 #define MPLP_GLF   0x10
 #define MPLP_NO_COMP 0x20
+#define MPLP_NO_ORPHAN 0x40
+#define MPLP_REALN   0x80
+#define MPLP_NO_HALFTRIM 0x100
 
 typedef struct {
 	int max_mq, min_mq, flag, min_baseQ;
@@ -463,19 +471,56 @@ typedef struct {
 typedef struct {
 	bamFile fp;
 	bam_iter_t iter;
-	int min_mq;
+	int min_mq, flag;
+	char *ref;
 } mplp_aux_t;
 
+typedef struct {
+	int n;
+	int *n_plp, *m_plp;
+	bam_pileup1_t **plp;
+} mplp_pileup_t;
+
 static int mplp_func(void *data, bam1_t *b)
 {
+	extern int bam_realn(bam1_t *b, const char *ref);
 	mplp_aux_t *ma = (mplp_aux_t*)data;
-	int ret;
+	int ret, cond = 0;
 	do {
+		cond = 0;
 		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);
+		if (ret < 0) break;
+		if (b->core.flag&BAM_FUNMAP) cond = 1;
+		else if (b->core.qual < ma->min_mq) cond = 1; 
+		else if ((ma->flag&MPLP_NO_ORPHAN) && (b->core.flag&1) && !(b->core.flag&2)) cond = 1;
+		if (ma->ref && !cond && (ma->flag&MPLP_REALN)) bam_realn(b, ma->ref);
+	} while (cond);
 	return ret;
 }
 
+static void group_smpl(mplp_pileup_t *m, bam_sample_t *sm, kstring_t *buf,
+					   int n, char *const*fn, int *n_plp, const bam_pileup1_t **plp)
+{
+	int i, j;
+	memset(m->n_plp, 0, m->n * sizeof(int));
+	for (i = 0; i < n; ++i) {
+		for (j = 0; j < n_plp[i]; ++j) {
+			const bam_pileup1_t *p = plp[i] + j;
+			uint8_t *q;
+			int id = -1;
+			q = bam_aux_get(p->b, "RG");
+			if (q) id = bam_smpl_rg2smid(sm, fn[i], (char*)q+1, buf);
+			if (id < 0) id = bam_smpl_rg2smid(sm, fn[i], 0, buf);
+			assert(id >= 0 && id < m->n);
+			if (m->n_plp[id] == m->m_plp[id]) {
+				m->m_plp[id] = m->m_plp[id]? m->m_plp[id]<<1 : 8;
+				m->plp[id] = realloc(m->plp[id], sizeof(bam_pileup1_t) * m->m_plp[id]);
+			}
+			m->plp[id][m->n_plp[id]++] = *p;
+		}
+	}
+}
+
 static int mpileup(mplp_conf_t *conf, int n, char **fn)
 {
 	mplp_aux_t **data;
@@ -492,18 +537,27 @@ static int mpileup(mplp_conf_t *conf, int n, char **fn)
 	bcf_t *bp = 0;
 	bcf_hdr_t *bh = 0;
 
+	bam_sample_t *sm = 0;
+	kstring_t buf;
+	mplp_pileup_t gplp;
+
+	memset(&gplp, 0, sizeof(mplp_pileup_t));
+	memset(&buf, 0, sizeof(kstring_t));
 	memset(&bc, 0, sizeof(bcf_call_t));
 	data = calloc(n, sizeof(void*));
 	plp = calloc(n, sizeof(void*));
 	n_plp = calloc(n, sizeof(int*));
+	sm = bam_smpl_init();
 
 	// read the header and initialize data
 	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]->flag = conf->flag;
 		data[i]->fp = bam_open(fn[i], "r");
 		h_tmp = bam_header_read(data[i]->fp);
+		bam_smpl_add(sm, fn[i], h_tmp->text);
 		if (conf->reg) {
 			int beg, end;
 			bam_index_t *idx;
@@ -526,6 +580,12 @@ static int mpileup(mplp_conf_t *conf, int n, char **fn)
 			bam_header_destroy(h_tmp);
 		}
 	}
+	gplp.n = sm->n;
+	gplp.n_plp = calloc(sm->n, sizeof(int));
+	gplp.m_plp = calloc(sm->n, sizeof(int));
+	gplp.plp = calloc(sm->n, sizeof(void*));
+
+	fprintf(stderr, "[%s] %d samples in %d input files\n", __func__, sm->n, n);
 	if (conf->fn_pos) hash = load_pos(conf->fn_pos, h);
 	// write the VCF header
 	if (conf->flag & MPLP_GLF) {
@@ -541,12 +601,8 @@ static int mpileup(mplp_conf_t *conf, int n, char **fn)
 		bh->name = malloc(s.l);
 		memcpy(bh->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);
+		for (i = 0; i < sm->n; ++i) {
+			kputs(sm->smpl[i], &s); kputc('\0', &s);
 		}
 		bh->l_smpl = s.l;
 		bh->sname = malloc(s.l);
@@ -555,11 +611,8 @@ static int mpileup(mplp_conf_t *conf, int n, char **fn)
 		free(s.s);
 		bcf_hdr_sync(bh);
 		bcf_hdr_write(bp, bh);
-	}
-	// mpileup
-	if (conf->flag & MPLP_GLF) {
 		bca = bcf_call_init(-1., conf->min_baseQ);
-		bcr = calloc(n, sizeof(bcf_callret1_t));
+		bcr = calloc(sm->n, sizeof(bcf_callret1_t));
 	}
 	ref_tid = -1; ref = 0;
 	iter = bam_mplp_init(n, mplp_func, (void**)data);
@@ -573,19 +626,20 @@ static int mpileup(mplp_conf_t *conf, int n, char **fn)
 		if (tid != ref_tid) {
 			free(ref);
 			if (conf->fai) ref = fai_fetch(conf->fai, h->target_name[tid], &ref_len);
+			for (i = 0; i < n; ++i) data[i]->ref = ref;
 			ref_tid = tid;
 		}
 		if (conf->flag & MPLP_GLF) {
 			int _ref0, ref16;
 			bcf1_t *b = calloc(1, sizeof(bcf1_t));
+			group_smpl(&gplp, sm, &buf, n, fn, n_plp, plp);
 			_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);
+			for (i = 0; i < gplp.n; ++i)
+				bcf_call_glfgen(gplp.n_plp[i], gplp.plp[i], ref16, bca, bcr + i);
+			bcf_call_combine(gplp.n, bcr, ref16, &bc);
 			bcf_call2bcf(tid, pos, &bc, b);
 			bcf_write(bp, bh, b);
-			//fprintf(stderr, "%d,%d,%d\n", b->tid, b->pos, b->l_str);
 			bcf_destroy(b);
 		} else {
 			printf("%s\t%d\t%c", h->target_name[tid], pos + 1, (ref && pos < ref_len)? ref[pos] : 'N');
@@ -608,7 +662,11 @@ static int mpileup(mplp_conf_t *conf, int n, char **fn)
 			putchar('\n');
 		}
 	}
+
 	bcf_close(bp);
+	bam_smpl_destroy(sm); free(buf.s);
+	for (i = 0; i < gplp.n; ++i) free(gplp.plp[i]);
+	free(gplp.plp); free(gplp.n_plp); free(gplp.m_plp);
 	if (hash) { // free the hash table
 		khint_t k;
 		for (k = kh_begin(hash); k < kh_end(hash); ++k)
@@ -635,7 +693,7 @@ int bam_mpileup(int argc, char *argv[])
 	mplp.max_mq = 60;
 	mplp.theta = 1e-3;
 	mplp.min_baseQ = 13;
-	while ((c = getopt(argc, argv, "gf:r:l:M:q:t:Q:u")) >= 0) {
+	while ((c = getopt(argc, argv, "gf:r:l:M:q:t:Q:uaORH")) >= 0) {
 		switch (c) {
 		case 't': mplp.theta = atof(optarg); break;
 		case 'f':
@@ -646,6 +704,10 @@ int bam_mpileup(int argc, char *argv[])
 		case 'l': mplp.fn_pos = strdup(optarg); break;
 		case 'g': mplp.flag |= MPLP_GLF; break;
 		case 'u': mplp.flag |= MPLP_NO_COMP; break;
+		case 'a': mplp.flag |= MPLP_NO_ORPHAN | MPLP_REALN | MPLP_NO_HALFTRIM; break;
+		case 'O': mplp.flag |= MPLP_NO_ORPHAN; break;
+		case 'H': mplp.flag |= MPLP_NO_HALFTRIM; break;
+		case 'R': mplp.flag |= MPLP_REALN; 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;
@@ -661,7 +723,8 @@ int bam_mpileup(int argc, char *argv[])
 		fprintf(stderr, "         -Q INT      min base quality [%d]\n", mplp.min_baseQ);
 		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, "         -g          generate GLF output\n");
+		fprintf(stderr, "         -g          generate BCF output\n");
+		fprintf(stderr, "         -u          do not compress BCF output\n");
 		fprintf(stderr, "\n");
 		fprintf(stderr, "Notes: Assuming diploid individuals.\n\n");
 		return 1;