* samtools-0.1.9-2 (r797)

 * mpileup: indel calling seems to be working

diff --git a/bam2bcf.c b/bam2bcf.c
index de6a3ed367b90bea3a4fc7430c9f4b4fa5bf26e4..be8cf1fe529a735800947b80b2a52c7fbaadb2d4 100644 (file)
--- a/bam2bcf.c
+++ b/bam2bcf.c
@@ -30,7 +30,7 @@ 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);
 }
 /* ref_base is the 4-bit representation of the reference base. It is
  * negative if we are looking at an indel. */
@@ -162,19 +162,38 @@ 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);
-       if (bc->ori_ref < 0) {
-               kputc('N', &s); kputc('\0', &s);
-               kputs("<INDEL>", &s); kputc('\0', &s);
-       } else {
+       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;
@@ -185,6 +204,7 @@ int bcf_call2bcf(int tid, int pos, bcf_call_t *bc, bcf1_t *b, bcf_callret1_t *bc
        }
        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);

diff --git a/bam2bcf.h b/bam2bcf.h
index e39ab7bfe90a6a6ec2a51f27e1325c32a1e4f2b5..570990e4174caec6220c130e0c343dec1d6dd562 100644 (file)
--- a/bam2bcf.h
+++ b/bam2bcf.h
@@ -13,6 +13,8 @@ typedef struct __bcf_callaux_t {
        // for internal uses
        int max_bases;
        int indel_types[4];
+       int maxins, indelreg;
+       char *inscns;
        uint16_t *bases;
        errmod_t *e;
 } bcf_callaux_t;
@@ -38,7 +40,8 @@ extern "C" {
        void bcf_call_destroy(bcf_callaux_t *bca);
        int bcf_call_glfgen(int _n, const bam_pileup1_t *pl, int ref_base, bcf_callaux_t *bca, bcf_callret1_t *r);
        int bcf_call_combine(int n, const bcf_callret1_t *calls, int ref_base /*4-bit*/, bcf_call_t *call);
-       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);
        int bcf_call_gap_prep(int n, int *n_plp, bam_pileup1_t **plp, int pos, bcf_callaux_t *bca, const char *ref);
 
 #ifdef __cplusplus

diff --git a/bam2bcf_indel.c b/bam2bcf_indel.c
index 1e8a2044305c4449713731fe95787b5105635d8e..25e2a86a72958ff19aa41e88b6b003816678029c 100644 (file)
--- a/bam2bcf_indel.c
+++ b/bam2bcf_indel.c
@@ -1,4 +1,5 @@
 #include <assert.h>
+#include <ctype.h>
 #include "bam.h"
 #include "bam2bcf.h"
 #include "ksort.h"
@@ -44,6 +45,19 @@ static inline int est_seqQ(const bcf_callaux_t *bca, int l, int l_run)
        return q < qh? q : qh;
 }
 
+static inline int est_indelreg(int pos, const char *ref, int l, char *ins4)
+{
+       int i, j, max = 0, max_i = pos, score = 0;
+       l = abs(l);
+       for (i = pos + 1, j = 0; ref[i]; ++i, ++j) {
+               if (ins4) score += (toupper(ref[i]) != "ACGTN"[(int)ins4[j%l]])? -10 : 1;
+               else score += (toupper(ref[i]) != toupper(ref[pos+1+j%l]))? -10 : 1;
+               if (score < 0) break;
+               if (max < score) max = score, max_i = i;
+       }
+       return max_i - pos;
+}
+
 int bcf_call_gap_prep(int n, int *n_plp, bam_pileup1_t **plp, int pos, bcf_callaux_t *bca, const char *ref)
 {
        extern void ks_introsort_uint32_t(int, uint32_t*);
@@ -149,10 +163,17 @@ int bcf_call_gap_prep(int n, int *n_plp, bam_pileup1_t **plp, int pos, bcf_calla
        ref2  = calloc(right - left + max_ins + 2, 1);
        query = calloc(right - left + max_rd_len + max_ins + 2, 1);
        score = calloc(N * n_types, sizeof(int));
+       bca->indelreg = 0;
        for (t = 0; t < n_types; ++t) {
-               int l;
+               int l, ir;
                ka_param2_t ap = ka_param2_qual;
                ap.band_width = abs(types[t]) + 3;
+               // compute indelreg
+               if (types[t] == 0) ir = 0;
+               else if (types[t] > 0) ir = est_indelreg(pos, ref, types[t], &inscns[t*max_ins]);
+               else ir = est_indelreg(pos, ref, -types[t], 0);
+               if (ir > bca->indelreg) bca->indelreg = ir;
+//             fprintf(stderr, "%d, %d, %d\n", pos, types[t], ir);
                // write ref2
                for (k = 0, j = left; j <= pos; ++j)
                        ref2[k++] = bam_nt16_nt4_table[bam_nt16_table[(int)ref[j]]];
@@ -233,7 +254,9 @@ int bcf_call_gap_prep(int n, int *n_plp, bam_pileup1_t **plp, int pos, bcf_calla
 //                             fprintf(stderr, "pos=%d read=%d:%d name=%s call=%d q=%d\n", pos, s, i, bam1_qname(p->b), types[sc[0]&0x3f], indelQ);
                        }
                }
-               // determine bca->indel_types[]
+               // determine bca->indel_types[] and bca->inscns
+               bca->maxins = max_ins;
+               bca->inscns = realloc(bca->inscns, bca->maxins * 4);
                for (t = 0; t < n_types; ++t)
                        sumq[t] = sumq[t]<<6 | t;
                for (t = 1; t < n_types; ++t) // insertion sort
@@ -247,8 +270,10 @@ int bcf_call_gap_prep(int n, int *n_plp, bam_pileup1_t **plp, int pos, bcf_calla
                        sumq[0] = tmp;
                }
                for (t = 0; t < 4; ++t) bca->indel_types[t] = B2B_INDEL_NULL;
-               for (t = 0; t < 4 && t < n_types; ++t)
+               for (t = 0; t < 4 && t < n_types; ++t) {
                        bca->indel_types[t] = types[sumq[t]&0x3f];
+                       memcpy(&bca->inscns[t * bca->maxins], &inscns[(sumq[t]&0x3f) * max_ins], bca->maxins);
+               }
                // update p->aux
                for (s = K = 0; s < n; ++s) {
                        for (i = 0; i < n_plp[s]; ++i, ++K) {
@@ -261,7 +286,6 @@ int bcf_call_gap_prep(int n, int *n_plp, bam_pileup1_t **plp, int pos, bcf_calla
                        }
                }               
        }
-       // FIXME: to set the inserted sequence
        free(score);
        // free
        free(types); free(inscns);

diff --git a/bam_plcmd.c b/bam_plcmd.c
index b56f879dfbf226e70bc07c7b2853705c6584f92f..54b8f91722b2f16ade955231e075764b3161d2d5 100644 (file)
--- a/bam_plcmd.c
+++ b/bam_plcmd.c
@@ -726,7 +726,8 @@ static int mpileup(mplp_conf_t *conf, int n, char **fn)
                        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, (conf->flag&(MPLP_FMT_DP|MPLP_FMT_SP))? bcr : 0, (conf->flag&MPLP_FMT_SP));
+                       bcf_call2bcf(tid, pos, &bc, b, (conf->flag&(MPLP_FMT_DP|MPLP_FMT_SP))? bcr : 0,
+                                                (conf->flag&MPLP_FMT_SP), 0, 0);
                        bcf_write(bp, bh, b);
                        bcf_destroy(b);
                        // call indels
@@ -736,7 +737,7 @@ static int mpileup(mplp_conf_t *conf, int n, char **fn)
                                bcf_call_combine(gplp.n, bcr, -1, &bc);
                                b = calloc(1, sizeof(bcf1_t));
                                bcf_call2bcf(tid, pos, &bc, b, (conf->flag&(MPLP_FMT_DP|MPLP_FMT_SP))? bcr : 0,
-                                                        (conf->flag&MPLP_FMT_SP));
+                                                        (conf->flag&MPLP_FMT_SP), bca, ref);
                                bcf_write(bp, bh, b);
                                bcf_destroy(b);
                        }

diff --git a/bamtk.c b/bamtk.c
index 4f881c04938d79c88e8c45fc4344a2dce14d02a9..743f33772f9de77edf9f7227a7d25a2fe1ecaf40 100644 (file)
--- a/bamtk.c
+++ b/bamtk.c
@@ -9,7 +9,7 @@
 #endif
 
 #ifndef PACKAGE_VERSION
-#define PACKAGE_VERSION "0.1.9-2 (r787)"
+#define PACKAGE_VERSION "0.1.9-3 (r797)"
 #endif
 
 int bam_taf2baf(int argc, char *argv[]);