prob_realn() seems working!

diff --git a/bam_md.c b/bam_md.c
index 746b2c03e6f471593721ce55f69c43ef04845923..7023046b87748de170de994a2565ec0b7a8172a2 100644 (file)
--- a/bam_md.c
+++ b/bam_md.c
@@ -261,6 +261,63 @@ int bam_realn(bam1_t *b, const char *ref)
        return 0;
 }
 
+int bam_prob_realn(bam1_t *b, const char *ref)
+{
+       int k, i, bw, x, y, yb, ye, xb, xe;
+       uint32_t *cigar = bam1_cigar(b);
+       bam1_core_t *c = &b->core;
+       ka_probpar_t conf = ka_probpar_def;
+       // find the start and end of the alignment
+       if (c->flag & BAM_FUNMAP) return -1;
+       x = c->pos, y = 0, yb = ye = xb = xe = -1;
+       for (k = 0; k < c->n_cigar; ++k) {
+               int op, l;
+               op = cigar[k]&0xf; l = cigar[k]>>4;
+               if (op == BAM_CMATCH) {
+                       if (yb < 0) yb = y;
+                       if (xb < 0) xb = x;
+                       ye = y + l; xe = x + l;
+                       x += l; y += l;
+               } else if (op == BAM_CSOFT_CLIP || op == BAM_CINS) y += l;
+               else if (op == BAM_CDEL) x += l;
+               else if (op == BAM_CREF_SKIP) return -1;
+       }
+       // set bandwidth and the start and the end
+       bw = 7;
+       if (abs((xe - xb) - (ye - yb)) > bw)
+               bw = abs((xe - xb) - (ye - yb)) + 3;
+       conf.bw = bw;
+       xb -= yb + bw/2; if (xb < 0) xb = 0;
+       xe += c->l_qseq - ye + bw/2;
+       if (xe - xb - c->l_qseq > bw)
+               xb += (xe - xb - c->l_qseq - bw) / 2, xe -= (xe - xb - c->l_qseq - bw) / 2;
+       { // glocal
+               uint8_t *s, *r, *q, *seq = bam1_seq(b), *qual = bam1_qual(b);
+               int *state;
+               s = calloc(c->l_qseq, 1);
+               for (i = 0; i < c->l_qseq; ++i) s[i] = bam_nt16_nt4_table[bam1_seqi(seq, i)];
+               r = calloc(xe - xb, 1);
+               for (i = xb; i < xe; ++i)
+                       r[i-xb] = bam_nt16_nt4_table[bam_nt16_table[(int)ref[i]]];
+               state = calloc(c->l_qseq, sizeof(int));
+               q = calloc(c->l_qseq, 1);
+               ka_prob_glocal(r, xe-xb, s, c->l_qseq, qual, &conf, state, q);
+               for (k = 0, x = c->pos, y = 0; k < c->n_cigar; ++k) {
+                       int op = cigar[k]&0xf, l = cigar[k]>>4;
+                       if (op == BAM_CMATCH) {
+                               for (i = y; i < y + l; ++i) {
+                                       if ((state[i]&3) != 0 || state[i]>>2 != x - xb + (i - y)) qual[i] = 0;
+                                       else qual[i] = qual[i] < q[i]? qual[i] : q[i];
+                               }
+                               x += l; y += l;
+                       } else if (op == BAM_CSOFT_CLIP || op == BAM_CINS) y += l;
+                       else if (op == BAM_CDEL) x += l;
+               }
+               free(s); free(r); free(q); free(state);
+       }
+       return 0;
+}
+
 int bam_fillmd(int argc, char *argv[])
 {
        int c, is_equal = 0, tid = -2, ret, len, is_bam_out, is_sam_in, is_uncompressed, max_nm = 0, is_realn, capQ = 0;
@@ -318,7 +375,8 @@ int bam_fillmd(int argc, char *argv[])
                                        fprintf(stderr, "[bam_fillmd] fail to find sequence '%s' in the reference.\n",
                                                        fp->header->target_name[tid]);
                        }
-                       if (is_realn) bam_realn(b, ref);
+//                     if (is_realn) bam_realn(b, ref);
+                       if (is_realn) bam_prob_realn(b, ref);
                        if (capQ > 10) {
                                int q = bam_cap_mapQ(b, ref, capQ);
                                if (b->core.qual > q) b->core.qual = q;

diff --git a/bam_plcmd.c b/bam_plcmd.c
index 358e9f99daea56c43dec513f82e64dacc13bd843..af43918f1ba0ea1d80ef86ecfbe75ac9e3553ad4 100644 (file)
--- a/bam_plcmd.c
+++ b/bam_plcmd.c
@@ -483,6 +483,7 @@ typedef struct {
 static int mplp_func(void *data, bam1_t *b)
 {
        extern int bam_realn(bam1_t *b, const char *ref);
+       extern int bam_prob_realn(bam1_t *b, const char *ref);
        extern int bam_cap_mapQ(bam1_t *b, char *ref, int thres);
        mplp_aux_t *ma = (mplp_aux_t*)data;
        int ret, skip = 0;
@@ -491,7 +492,7 @@ static int mplp_func(void *data, bam1_t *b)
                ret = ma->iter? bam_iter_read(ma->fp, ma->iter, b) : bam_read1(ma->fp, b);
                if (ret < 0) break;
                skip = 0;
-               if (has_ref && (ma->flag&MPLP_REALN)) bam_realn(b, ma->ref);
+               if (has_ref && (ma->flag&MPLP_REALN)) bam_prob_realn(b, ma->ref);
                if (has_ref && ma->capQ_thres > 10) {
                        int q = bam_cap_mapQ(b, ma->ref, ma->capQ_thres);
                        if (q < 0) skip = 1;

diff --git a/kaln.c b/kaln.c
index e25ebb5d1ac3d2e67c96db2ad266beaf5602bde1..d8edb91f13ef1dc8aabd0b066068b7a6c8b98baa 100644 (file)
--- a/kaln.c
+++ b/kaln.c
@@ -371,9 +371,9 @@ uint32_t *ka_global_core(uint8_t *seq1, int len1, uint8_t *seq2, int len2, const
        return cigar;
 }
 
-/************************
- * Probabilistic glocal *
- ************************/
+/*****************************************
+ * Probabilistic banded glocal alignment *
+ *****************************************/
 
 static float g_qual2prob[256];
 
@@ -495,7 +495,7 @@ int ka_prob_glocal(const uint8_t *_ref, int l_ref, const uint8_t *_query, int l_
        // b[l_query-1..1]
        for (i = l_query - 1; i >= 1; --i) {
                int beg = 1, end = l_ref, x, _beg, _end;
-               double *bi = b[i], *bi1 = b[i+1];
+               double *bi = b[i], *bi1 = b[i+1], y = (i > 1);
                x = i - bw; beg = beg > x? beg : x;
                x = i + bw; end = end < x? end : x;
                for (k = end; k >= beg; --k) {
@@ -507,12 +507,13 @@ int ka_prob_glocal(const uint8_t *_ref, int l_ref, const uint8_t *_query, int l_
                        bi[u+0] = e * m[0] * bi1[v11+0] + .25 * m[1] * bi1[v10+1] + m[2] * bi[v01+2];
                        bi[u+1] = e * m[3] * bi1[v11+0] + .25 * m[4] * bi1[v10+1];
                        // FIXME: I do not know why I need this (i>1) factor, but only with it the result makes sense.
-                       bi[u+2] = (e * m[6] * bi1[v11+0] + m[8] * bi[v01+2]) * (i > 1);
+                       bi[u+2] = (e * m[6] * bi1[v11+0] + m[8] * bi[v01+2]) * y;
 //                     fprintf(stderr, "B (%d,%d;%d): %lg,%lg,%lg\n", i, k, u, bi[u], bi[u+1], bi[u+2]); // DEBUG
                }
                // rescale
                set_u(_beg, bw, i, beg); set_u(_end, bw, i, end); _end += 2;
-               for (k = _beg; k <= _end; ++k) bi[k] /= s[i];
+               y = s[i];
+               for (k = _beg; k <= _end; ++k) bi[k] /= y;
        }
        { // b[0]
                int beg = 1, end = l_ref < bw + 1? l_ref : bw + 1;
@@ -531,21 +532,21 @@ int ka_prob_glocal(const uint8_t *_ref, int l_ref, const uint8_t *_query, int l_
        /*** MAP ***/
        for (i = 1; i <= l_query; ++i) {
                double sum = 0., *fi = f[i], *bi = b[i], max = 0.;
-               int beg = 0, end = l_ref, x, max_k = -1;
+               int beg = 1, end = l_ref, x, max_k = -1;
                x = i - bw; beg = beg > x? beg : x;
                x = i + bw; end = end < x? end : x;
                for (k = beg; k <= end; ++k) {
                        int u;
                        double z;
                        set_u(u, bw, i, k);
-                       z = fi[u+0] * bi[u+0]; if (z > max) max = z, max_k = k<<2 | 0; sum += z;
-                       z = fi[u+1] * bi[u+1]; if (z > max) max = z, max_k = k<<2 | 1; sum += z;
+                       z = fi[u+0] * bi[u+0]; if (z > max) max = z, max_k = (k-1)<<2 | 0; sum += z;
+                       z = fi[u+1] * bi[u+1]; if (z > max) max = z, max_k = (k-1)<<2 | 1; sum += z;
                }
                max /= sum; sum *= s[i]; // if everything works as is expected, sum == 1.0
                if (state) state[i-1] = max_k;
-               if (q) q[i-1] = -4.343 * log(1. - max);
+               if (q) k = -4.343 * log(1. - max), q[i-1] = k > 100? 99 : k;
 #ifdef _MAIN
-               fprintf(stderr, "(%.10lg,%.10lg) (%d,%d:%d)~%lg\n", pb, sum, i, max_k>>2, max_k&3, max); // DEBUG
+               fprintf(stderr, "(%.10lg,%.10lg) (%d,%d:%d)~%lg\n", pb, sum, i-1, max_k>>2, max_k&3, max); // DEBUG
 #endif
        }
        /*** free ***/