X-Git-Url: https://git.donarmstrong.com/?a=blobdiff_plain;f=bam2bcf.c;h=c2d994479a9094ea330291db34290e15faca8850;hb=0917f66462dc2a0e7be6a33cb4f3e969e3bb1c9b;hp=7fe178a4dd66a780c2f5385620c669b49bcb177d;hpb=6dc53c5ab2964e68e8936850d2339d047d2433ef;p=samtools.git

diff --git a/bam2bcf.c b/bam2bcf.c
index 7fe178a..c2d9944 100644
--- a/bam2bcf.c
+++ b/bam2bcf.c
@@ -6,19 +6,13 @@
 #include "errmod.h"
 #include "bcftools/bcf.h"
 
-#define END_DIST_THRES 11
-
 extern	void ks_introsort_uint32_t(size_t n, uint32_t a[]);
 
 #define CALL_ETA 0.03f
 #define CALL_MAX 256
 #define CALL_DEFTHETA 0.83f
 
-struct __bcf_callaux_t {
-	int max_bases, capQ, min_baseQ;
-	uint16_t *bases;
-	errmod_t *e;
-};
+#define CAP_DIST 25
 
 bcf_callaux_t *bcf_call_init(double theta, int min_baseQ)
 {
@@ -26,6 +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 = 80;
 	bca->min_baseQ = min_baseQ;
 	bca->e = errmod_init(1. - theta);
 	return bca;
@@ -35,16 +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, k, 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;
@@ -52,33 +50,39 @@ int bcf_call_glfgen(int _n, const bam_pileup1_t *pl, int ref_base /*4-bit*/, bcf
 		bca->bases = (uint16_t*)realloc(bca->bases, 2 * bca->max_bases);
 	}
 	// fill the bases array
-	memset(r->qsum, 0, 4 * sizeof(float));
-	for (i = n = 0, r->sum_Q2 = 0; i < _n; ++i) {
+	memset(r, 0, sizeof(bcf_callret1_t));
+	for (i = n = 0; i < _n; ++i) {
 		const bam_pileup1_t *p = pl + i;
-		int q, b, mapQ;
-		int min_dist;
+		int q, b, mapQ, baseQ, is_diff, min_dist;
 		// set base
-		if (p->is_del || (p->b->core.flag&BAM_FUNMAP)) continue; // skip unmapped reads and deleted bases
-		q = (int)bam1_qual(p->b)[p->qpos]; // base quality
+		if (p->is_del || p->is_refskip || (p->b->core.flag&BAM_FUNMAP)) continue;
+		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;
-		r->sum_Q2 += mapQ * mapQ;
 		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 other information
+		// collect annotations
 		r->qsum[b] += q;
-		k = (ref4 < 4 && b == ref4)? 0 : 1;
-		k = k<<1 | bam1_strand(p->b);
-		++r->d[k];
-		// calculate min_dist
+		++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;
-		k = (k&2) | (min_dist <= END_DIST_THRES);
-		++r->ed[k];
+		if (min_dist > CAP_DIST) min_dist = CAP_DIST;
+		r->anno[1<<2|is_diff<<1|0] += baseQ;
+		r->anno[1<<2|is_diff<<1|1] += baseQ * baseQ;
+		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
@@ -90,9 +94,11 @@ int bcf_call_combine(int n, const bcf_callret1_t *calls, int ref_base /*4-bit*/,
 {
 	int ref4, i, j, qsum[4];
 	int64_t tmp;
-	call->ori_ref = ref4 = bam_nt16_nt4_table[ref_base];
-	if (ref4 > 4) ref4 = 4;
-	// calculate esum
+	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)
 		for (j = 0; j < 4; ++j)
@@ -112,9 +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;
@@ -142,52 +150,98 @@ 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
-	memset(call->d, 0, 4 * sizeof(int));
-	memset(call->ed, 0, 4 * sizeof(int));
+	memset(call->anno, 0, 16 * sizeof(int));
 	for (i = call->depth = 0, tmp = 0; i < n; ++i) {
 		call->depth += calls[i].depth;
-		for (j = 0; j < 4; ++j) call->d[j] += calls[i].d[j], call->ed[j] += calls[i].ed[j];
-		tmp += calls[i].sum_Q2;
+		for (j = 0; j < 16; ++j) call->anno[j] += calls[i].anno[j];
 	}
-	call->rmsQ = (int)(sqrt((double)tmp / call->depth) + .499);
 	return 0;
 }
 
-int bcf_call2bcf(int tid, int pos, bcf_call_t *bc, bcf1_t *b)
+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; ++j)
+					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
-	kputs("MQ=", &s); kputw(bc->rmsQ, &s);
-	kputs(";DP4=", &s);
-	for (i = 0; i < 4; ++i) {
-		if (i) kputc(',', &s);
-		kputw(bc->d[i], &s);
-	}
-	kputs(";ED4=", &s);
-	for (i = 0; i < 4; ++i) {
+	if (bc->ori_ref < 0) kputs("INDEL;", &s);
+	kputs("I16=", &s);
+	for (i = 0; i < 16; ++i) {
 		if (i) kputc(',', &s);
-		kputw(bc->ed[i], &s);
+		kputw(bc->anno[i], &s);
 	}
 	kputc('\0', &s);
 	// FMT
-	kputs("PL", &s); kputc('\0', &s);
+	kputs("PL", &s);
+	if (bcr) {
+		kputs(":DP", &s);
+		if (is_SP) kputs(":SP", &s);
+	}
+	kputc('\0', &s);
 	b->m_str = s.m; b->str = s.s; b->l_str = s.l;
-	bcf_sync(bc->n, b);
+	bcf_sync(b);
 	memcpy(b->gi[0].data, bc->PL, b->gi[0].len * bc->n);
+	if (bcr) {
+		uint16_t *dp = (uint16_t*)b->gi[1].data;
+		uint8_t *sp = is_SP? b->gi[2].data : 0;
+		for (i = 0; i < bc->n; ++i) {
+			bcf_callret1_t *p = bcr + i;
+			dp[i] = p->depth < 0xffff? p->depth : 0xffff;
+			if (is_SP) {
+				if (p->anno[0] + p->anno[1] < 2 || p->anno[2] + p->anno[3] < 2
+					|| p->anno[0] + p->anno[2] < 2 || p->anno[1] + p->anno[3] < 2)
+				{
+					sp[i] = 0;
+				} else {
+					double left, right, two;
+					int x;
+					kt_fisher_exact(p->anno[0], p->anno[1], p->anno[2], p->anno[3], &left, &right, &two);
+					x = (int)(-4.343 * log(two) + .499);
+					if (x > 255) x = 255;
+					sp[i] = x;
+				}
+			}
+		}
+	}
 	return 0;
 }