X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=bam2bcf.c;h=be8cf1fe529a735800947b80b2a52c7fbaadb2d4;hb=599bea10bc3487a0cbdf505e792a21a13fee2b11;hp=d5a8dd440286cc2b6d0b119df6e6940d8ea05bbf;hpb=98946f7c4842bc7c0429bb78982366ce46c70624;p=samtools.git

diff --git a/bam2bcf.c b/bam2bcf.c
index d5a8dd4..be8cf1f 100644
--- a/bam2bcf.c
+++ b/bam2bcf.c
@@ -3,6 +3,7 @@
 #include "bam.h"
 #include "kstring.h"
 #include "bam2bcf.h"
+#include "errmod.h"
 #include "bcftools/bcf.h"
 
 extern	void ks_introsort_uint32_t(size_t n, uint32_t a[]);
@@ -19,9 +20,7 @@ bcf_callaux_t *bcf_call_init(double theta, int min_baseQ)
 	if (theta <= 0.) theta = CALL_DEFTHETA;
 	bca = calloc(1, sizeof(bcf_callaux_t));
 	bca->capQ = 60;
-	bca->openQ = 40;
-	bca->extQ = 20;
-	bca->tandemQ = 100;
+	bca->openQ = 40; bca->extQ = 20; bca->tandemQ = 80;
 	bca->min_baseQ = min_baseQ;
 	bca->e = errmod_init(1. - theta);
 	return bca;
@@ -31,16 +30,19 @@ void bcf_call_destroy(bcf_callaux_t *bca)
 {
 	if (bca == 0) return;
 	errmod_destroy(bca->e);
-	free(bca->bases); free(bca);
+	free(bca->bases); free(bca->inscns); free(bca);
 }
-
-int bcf_call_glfgen(int _n, const bam_pileup1_t *pl, int ref_base /*4-bit*/, bcf_callaux_t *bca, bcf_callret1_t *r)
+/* ref_base is the 4-bit representation of the reference base. It is
+ * negative if we are looking at an indel. */
+int bcf_call_glfgen(int _n, const bam_pileup1_t *pl, int ref_base, bcf_callaux_t *bca, bcf_callret1_t *r)
 {
-	int i, n, ref4;
+	int i, n, ref4, is_indel;
 	memset(r, 0, sizeof(bcf_callret1_t));
-	ref4 = bam_nt16_nt4_table[ref_base];
+	if (ref_base >= 0) {
+		ref4 = bam_nt16_nt4_table[ref_base];
+		is_indel = 0;
+	} else ref4 = 4, is_indel = 1;
 	if (_n == 0) return -1;
-
 	// enlarge the bases array if necessary
 	if (bca->max_bases < _n) {
 		bca->max_bases = _n;
@@ -54,18 +56,23 @@ int bcf_call_glfgen(int _n, const bam_pileup1_t *pl, int ref_base /*4-bit*/, bcf
 		int q, b, mapQ, baseQ, is_diff, min_dist;
 		// set base
 		if (p->is_del || p->is_refskip || (p->b->core.flag&BAM_FUNMAP)) continue;
-		baseQ = q = (int)bam1_qual(p->b)[p->qpos]; // base quality
+		baseQ = q = is_indel? p->aux&0xff : (int)bam1_qual(p->b)[p->qpos]; // base/indel quality
 		if (q < bca->min_baseQ) continue;
 		mapQ = p->b->core.qual < bca->capQ? p->b->core.qual : bca->capQ;
 		if (q > mapQ) q = mapQ;
 		if (q > 63) q = 63;
 		if (q < 4) q = 4;
-		b = bam1_seqi(bam1_seq(p->b), p->qpos); // base
-		b = bam_nt16_nt4_table[b? b : ref_base]; // b is the 2-bit base
+		if (!is_indel) {
+			b = bam1_seqi(bam1_seq(p->b), p->qpos); // base
+			b = bam_nt16_nt4_table[b? b : ref_base]; // b is the 2-bit base
+			is_diff = (ref4 < 4 && b == ref4)? 0 : 1;
+		} else {
+			b = p->aux>>8&0x3f;
+			is_diff = (b != 0);
+		}
 		bca->bases[n++] = q<<5 | (int)bam1_strand(p->b)<<4 | b;
 		// collect annotations
 		r->qsum[b] += q;
-		is_diff = (ref4 < 4 && b == ref4)? 0 : 1;
 		++r->anno[0<<2|is_diff<<1|bam1_strand(p->b)];
 		min_dist = p->b->core.l_qseq - 1 - p->qpos;
 		if (min_dist > p->qpos) min_dist = p->qpos;
@@ -83,115 +90,14 @@ int bcf_call_glfgen(int _n, const bam_pileup1_t *pl, int ref_base /*4-bit*/, bcf
 	return r->depth;
 }
 
-int bcf_call_glfgen_gap(int pos, int _n, const bam_pileup1_t *pl, bcf_callaux_t *bca, bcf_callret1_t *r)
-{
-	int i, n, n_ins, n_del;
-	memset(r, 0, sizeof(bcf_callret1_t));
-	if (_n == 0) return -1;
-
-	// enlarge the bases array if necessary
-	if (bca->max_bases < _n) {
-		bca->max_bases = _n;
-		kroundup32(bca->max_bases);
-		bca->bases = (uint16_t*)realloc(bca->bases, 2 * bca->max_bases);
-	}
-	// fill the bases array
-	memset(r, 0, sizeof(bcf_callret1_t));
-	r->indelreg = 10000;
-	for (i = n = 0; i < _n; ++i) {
-		const bam_pileup1_t *p = pl + i;
-		int q, b, mapQ, indelQ, is_diff, min_dist;
-		if (p->is_del || p->is_refskip || (p->b->core.flag&BAM_FUNMAP)) continue;
-		{ // compute indel (base) quality
-			// this can be made more efficient, but realignment is the bottleneck anyway
-			int j, k, x, y, op, len = 0, max_left, max_rght, seqQ, indelreg;
-			bam1_core_t *c = &p->b->core;
-			uint32_t *cigar = bam1_cigar(p->b);
-			uint8_t *qual = bam1_qual(p->b);
-			for (k = y = 0, x = c->pos; k < c->n_cigar && y <= p->qpos; ++k) {
-				op = cigar[k]&0xf;
-				len = cigar[k]>>4;
-				if (op == BAM_CMATCH) {
-					if (pos > x && pos < x + len) break;
-					x += len; y += len;
-				} else if (op == BAM_CINS || op == BAM_CSOFT_CLIP) y += len;
-				else if (op == BAM_CDEL || op == BAM_CREF_SKIP) x += len;
-			}
-			if (k == c->n_cigar) continue; // this actually should not happen
-			max_left = max_rght = 0; indelreg = 0;
-			if (pos == x + len - 1 && k+2 < c->n_cigar && ((cigar[k+1]&0xf) == BAM_CINS || (cigar[k+1]&0xf) == BAM_CDEL)
-				&& (cigar[k+2]&0xf) == BAM_CMATCH)
-			{
-				for (j = y; j < y + len; ++j)
-					if (max_left < qual[j]) max_left = qual[j];
-				if ((cigar[k+1]&0xf) == BAM_CINS) y += cigar[k+1]>>4;
-				else x += cigar[k+1]>>4;
-				op = cigar[k+2]&0xf; len = cigar[k+2]>>4;
-				for (j = y; j < y + len; ++j) {
-					if (max_rght < qual[j]) max_rght = qual[j];
-					if (qual[j] > BAM2BCF_INDELREG_THRES && indelreg == 0)
-						indelreg = j - y + 1;
-				}
-			} else {
-				for (j = y; j <= p->qpos; ++j)
-					if (max_left < qual[j]) max_left = qual[j];
-				for (j = p->qpos + 1; j < y + len; ++j)
-					if (max_rght < qual[j]) max_rght = qual[j];
-					
-			}
-			indelQ = max_left < max_rght? max_left : max_rght;
-			// estimate the sequencing error rate
-			seqQ = bca->openQ;
-			if (p->indel != 0) seqQ += bca->extQ * (abs(p->indel) - 1); // FIXME: better to model homopolymer
-			if (p->indel != 0) { // a different model for tandem repeats
-				uint8_t *seq = bam1_seq(p->b);
-				int tandemQ, qb = bam1_seqi(seq, p->qpos), l;
-				for (j = p->qpos + 1; j < c->l_qseq; ++j)
-					if (qb != bam1_seqi(seq, j)) break;
-				l = j;
-				for (j = (int)p->qpos - 1; j >= 0; --j)
-					if (qb != bam1_seqi(seq, j)) break;
-				l = l - (j + 1);
-				tandemQ = (int)((double)(abs(p->indel)) / l * bca->tandemQ + .499);
-				if (seqQ > tandemQ) seqQ = tandemQ;
-			}
-//			fprintf(stderr, "%s\t%d\t%d\t%d\t%d\t%d\t%d\n", bam1_qname(p->b), pos+1, p->indel, indelQ, seqQ, max_left, max_rght);
-			if (indelQ > seqQ) indelQ = seqQ;
-			q = indelQ;
-		}
-		if (q < bca->min_baseQ) continue;
-		mapQ = p->b->core.qual < bca->capQ? p->b->core.qual : bca->capQ;
-		if (q > mapQ) q = mapQ;
-		if (q > 63) q = 63;
-		if (q < 4) q = 4;
-		b = p->indel? 1 : 0;
-		bca->bases[n++] = q<<5 | (int)bam1_strand(p->b)<<4 | b;
-		// collect annotations
-		r->qsum[b] += q;
-		is_diff = b;
-		++r->anno[0<<2|is_diff<<1|bam1_strand(p->b)];
-		min_dist = p->b->core.l_qseq - 1 - p->qpos;
-		if (min_dist > p->qpos) min_dist = p->qpos;
-		if (min_dist > CAP_DIST) min_dist = CAP_DIST;
-		r->anno[1<<2|is_diff<<1|0] += indelQ;
-		r->anno[1<<2|is_diff<<1|1] += indelQ * indelQ;
-		r->anno[2<<2|is_diff<<1|0] += mapQ;
-		r->anno[2<<2|is_diff<<1|1] += mapQ * mapQ;
-		r->anno[3<<2|is_diff<<1|0] += min_dist;
-		r->anno[3<<2|is_diff<<1|1] += min_dist * min_dist;
-	}
-	r->depth = n;
-	// glfgen
-	errmod_cal(bca->e, n, 2, bca->bases, r->p);
-	return r->depth;	
-}
-
 int bcf_call_combine(int n, const bcf_callret1_t *calls, int ref_base /*4-bit*/, bcf_call_t *call)
 {
 	int ref4, i, j, qsum[4];
 	int64_t tmp;
-	call->ori_ref = ref4 = bam_nt16_nt4_table[ref_base];
-	if (ref4 > 4) ref4 = 4;
+	if (ref_base >= 0) {
+		call->ori_ref = ref4 = bam_nt16_nt4_table[ref_base];
+		if (ref4 > 4) ref4 = 4;
+	} else call->ori_ref = -1, ref4 = 0;
 	// calculate qsum
 	memset(qsum, 0, 4 * sizeof(int));
 	for (i = 0; i < n; ++i)
@@ -212,9 +118,11 @@ int bcf_call_combine(int n, const bcf_callret1_t *calls, int ref_base /*4-bit*/,
 			else break;
 		}
 	}
-	if (((ref4 < 4 && j < 4) || (ref4 == 4 && j < 5)) && i >= 0)
-		call->unseen = j, call->a[j++] = qsum[i]&3;
-	call->n_alleles = j;
+	if (ref_base >= 0) { // for SNPs, find the "unseen" base
+		if (((ref4 < 4 && j < 4) || (ref4 == 4 && j < 5)) && i >= 0)
+			call->unseen = j, call->a[j++] = qsum[i]&3;
+		call->n_alleles = j;
+	} else call->n_alleles = j;
 	// set the PL array
 	if (call->n < n) {
 		call->n = n;
@@ -242,6 +150,7 @@ int bcf_call_combine(int n, const bcf_callret1_t *calls, int ref_base /*4-bit*/,
 				PL[j] = y;
 			}
 		}
+//		if (ref_base < 0) fprintf(stderr, "%d,%d,%f,%d\n", call->n_alleles, x, sum_min, call->unseen);
 		call->shift = (int)(sum_min + .499);
 	}
 	// combine annotations
@@ -253,24 +162,49 @@ int bcf_call_combine(int n, const bcf_callret1_t *calls, int ref_base /*4-bit*/,
 	return 0;
 }
 
-int bcf_call2bcf(int tid, int pos, bcf_call_t *bc, bcf1_t *b, bcf_callret1_t *bcr, int is_SP)
+int bcf_call2bcf(int tid, int pos, bcf_call_t *bc, bcf1_t *b, bcf_callret1_t *bcr, int is_SP,
+				 const bcf_callaux_t *bca, const char *ref)
 {
 	extern double kt_fisher_exact(int n11, int n12, int n21, int n22, double *_left, double *_right, double *two);
 	kstring_t s;
-	int i;
+	int i, j;
 	b->n_smpl = bc->n;
 	b->tid = tid; b->pos = pos; b->qual = 0;
 	s.s = b->str; s.m = b->m_str; s.l = 0;
 	kputc('\0', &s);
-	kputc("ACGTN"[bc->ori_ref], &s); kputc('\0', &s);
-	for (i = 1; i < 5; ++i) {
-		if (bc->a[i] < 0) break;
-		if (i > 1) kputc(',', &s);
-		kputc(bc->unseen == i? 'X' : "ACGT"[bc->a[i]], &s);
+	if (bc->ori_ref < 0) { // an indel
+		// write REF
+		kputc(ref[pos], &s);
+		for (j = 0; j < bca->indelreg; ++j) kputc(ref[pos+1+j], &s);
+		kputc('\0', &s);
+		// write ALT
+		kputc(ref[pos], &s);
+		for (i = 1; i < 4; ++i) {
+			if (bc->a[i] < 0) break;
+			if (i > 1) kputc(',', &s);
+			if (bca->indel_types[bc->a[i]] < 0) { // deletion
+				for (j = -bca->indel_types[bc->a[i]]; j < bca->indelreg; ++i)
+					kputc(ref[pos+1+j], &s);
+			} else { // insertion; cannot be a reference unless a bug
+				char *inscns = &bca->inscns[bc->a[i] * bca->maxins];
+				for (j = 0; j < bca->indel_types[bc->a[i]]; ++j)
+					kputc("ACGTN"[(int)inscns[j]], &s);
+				for (j = 0; j < bca->indelreg; ++j) kputc(ref[pos+1+j], &s);
+			}
+		}
+		kputc('\0', &s);
+	} else { // a SNP
+		kputc("ACGTN"[bc->ori_ref], &s); kputc('\0', &s);
+		for (i = 1; i < 5; ++i) {
+			if (bc->a[i] < 0) break;
+			if (i > 1) kputc(',', &s);
+			kputc(bc->unseen == i? 'X' : "ACGT"[bc->a[i]], &s);
+		}
+		kputc('\0', &s);
 	}
 	kputc('\0', &s);
-	kputc('\0', &s);
 	// INFO
+	if (bc->ori_ref < 0) kputs("INDEL;", &s);
 	kputs("I16=", &s);
 	for (i = 0; i < 16; ++i) {
 		if (i) kputc(',', &s);