ViewVC Help
View File | Revision Log | Show Annotations | View Changeset | Root Listing
root/gclib/gclib/GAlnExtend.cpp
(Generate patch)
# Line 53 | Line 53
53    if (flast_d[d-1][diag-1] > GMAX(flast_d[d-1][diag], flast_d[d-1][diag+1])) {
54        *row1 = flast_d[d-1][diag-1];
55        return diag-1;
56 <      }
56 >      }
57    if (flast_d[d-1][diag] > flast_d[d-1][diag+1]) {
58        *row1 = flast_d[d-1][diag];
59        return diag;
# Line 65 | Line 65
65   void GapXEditScript::print() { //debug
66        GapXEditScript* p=this;
67        do {
68 <        GMessage("%d%c ",p->num, xgapcodes[p->op_type]);
69 <        } while ((p=p->next)!=NULL);
68 >        GMessage("%d%c ",p->num, xgapcodes[p->op_type]);
69 >        } while ((p=p->next)!=NULL);
70        GMessage("\n");
71      }
72  
# Line 93 | Line 93
93      else
94          g = BLAST_Gcd(*a, BLAST_Gcd(*b, *c));
95      if (g > 1) {
96 <        *a /= g;
96 >                *a /= g;
97          *b /= g;
98          *c /= g;
99      }
# Line 101 | Line 101
101   }
102  
103  
104 + uint16 get6mer(char* p) {
105 +  uint16 r=gdna2bit(p,3);
106 +  r <<= 6;
107 +  r |= gdna2bit(p,3);
108 +  return r;
109 + }
110 +
111 +
112 + void table6mers(const char* s, int slen, GVec<uint16>* amers[]) {
113 + for (uint16 i=0; i < slen; i++) {
114 +   char* p = (char*)(s+i);
115 +   uint16 v=get6mer(p);
116 +   if (amers[v]==NULL) {
117 +      amers[v]=new GVec<uint16>(1);
118 +      }
119 +   amers[v]->Add(i);
120 + }
121 + }
122 +
123 + GVec<uint16>* match6mer(char* start, GVec<uint16>* amers[]) {
124 +  //careful: this is broken if start+5 falls beyond the end of the string!
125 +  uint16 r=get6mer(start);
126 +  return amers[r];
127 + }
128  
129   //signal that a diagonal is invalid
130   static const int kInvalidOffset = -2;
# Line 129 | Line 153
153               //    }
154        }
155      else { //forward lookup
156 <             while (seq1_index < len1 && seq2_index < len2 &&
156 >             while (seq1_index < len1 && seq2_index < len2 &&
157                     //seq1[seq1_index] < 4 &&
158                     seq1[seq1_index] == seq2[seq2_index]) {
159                  ++seq1_index;
# Line 204 | Line 228
228      int longest_match_run;
229      bool end1_reached, end2_reached;
230      GXMemPool* mem_pool;
231 <
231 >
232      /* ordinary dynamic programming alignment, for each offset
233         in seq1, walks through offsets in seq2 until an X-dropoff
234 <       test fails, saving the best score encountered along
234 >       test fails, saving the best score encountered along
235         the way. Instead of score, this code tracks the 'distance'
236         (number of mismatches plus number of gaps) between seq1
237         and seq2. Instead of walking through sequence offsets, it
238         walks through diagonals that can achieve a given distance.
239 <    
239 >
240         Note that in what follows, the numbering of diagonals implies
241 <       a dot matrix where increasing seq1 offsets go to the right on
241 >       a dot matrix where increasing seq1 offsets go to the right on
242         the x axis, and increasing seq2 offsets go up the y axis.
243         The gapped alignment thus proceeds up and to the right in
244         the graph, and diagonals are numbered increasing to the right */
# Line 223 | Line 247
247      best_diag = 0;
248  
249      /* set the number of distinct distances the algorithm will
250 <       examine in the search for an optimal alignment. The
251 <       heuristic worst-case running time of the algorithm is
250 >       examine in the search for an optimal alignment. The
251 >       heuristic worst-case running time of the algorithm is
252         O(max_dist**2 + (len1+len2)); for sequences which are
253         very similar, the average running time will be sig-
254         nificantly better than this */
255  
256      max_dist = GMIN(GREEDY_MAX_COST,
257 <                   len2 / GREEDY_MAX_COST_FRACTION + 1);
257 >                   (len2/GREEDY_MAX_COST_FRACTION + 1));
258  
259      /* the main loop assumes that the index of all diagonals is
260 <       biased to lie in the middle of allocated bookkeeping
260 >       biased to lie in the middle of allocated bookkeeping
261         structures */
262  
263      diag_origin = max_dist + 2;
# Line 251 | Line 275
275         xdrop_offset gives the distance backwards in the score
276         array to look */
277  
278 <    xdrop_offset = (xdrop_threshold + match_cost / 2) /
278 >    xdrop_offset = (xdrop_threshold + match_cost / 2) /
279                             (match_cost + mismatch_cost) + 1;
280 <    
280 >
281      // find the offset of the first mismatch between seq1 and seq2
282  
283      index = s_FindFirstMismatch(seq1, len1,  seq2, len2, 0, 0, reverse);
# Line 277 | Line 301
301          if (edit_block != NULL)
302              //GapPrelimEditBlockAdd(edit_block, eGapAlignSub, index);
303              edit_block->opRep(index);
304 <        return 0;
304 >        return 0;
305      }
306  
307      // set up the memory pool
# Line 287 | Line 311
311        }
312      else if (mem_pool == NULL) {
313         aux_data.space = mem_pool = new GXMemPool();
314 <    }
314 >    }
315      else {
316          mem_pool->refresh();
317      }
318 <    
318 >
319      /* set up the array of per-distance maximum scores. There
320         are max_diags + xdrop_offset distances to track, the first
321         xdrop_offset of which are 0 */
# Line 299 | Line 323
323      max_score = aux_data.max_score + xdrop_offset;
324      for (index = 0; index < xdrop_offset; index++)
325          aux_data.max_score[index] = 0;
326 <    
326 >
327      // fill in the initial offsets of the distance matrix
328  
329      last_seq2_off[0][diag_origin] = seq1_index;
# Line 328 | Line 352
352  
353          // compute the score for distance d corresponding to the X-dropoff criterion
354  
355 <        xdrop_score = max_score[d - xdrop_offset] +
355 >        xdrop_score = max_score[d - xdrop_offset] +
356                        (match_cost + mismatch_cost) * d - xdrop_threshold;
357 <        xdrop_score = (int)ceil((double)xdrop_score / (match_cost / 2));
357 >        xdrop_score = (int)ceil((double)xdrop_score / (match_cost / 2));
358  
359          // for each diagonal of interest
360          for (k = tmp_diag_lower; k <= tmp_diag_upper; k++) {
361              /* find the largest offset into seq2 that increases
362                 the distance from d-1 to d (i.e. keeps the alignment
363 <               from getting worse for as long as possible), then
363 >               from getting worse for as long as possible), then
364                 choose the offset into seq1 that will keep the
365 <               resulting diagonal fixed at k
366 <            
365 >               resulting diagonal fixed at k
366 >
367                 Note that this requires kInvalidOffset+1 to be smaller
368                 than any valid offset into seq2, i.e. to be negative */
369  
# Line 360 | Line 384
384                  continue;
385              }
386              diag_upper = k;
387 <            
388 <            /* slide down diagonal k until a mismatch
387 >
388 >            /* slide down diagonal k until a mismatch
389                 occurs. As long as only matches are encountered,
390                 the current distance d will not change */
391  
# Line 397 | Line 421
421                 can each be of size at most max_diags+2 */
422  
423              if (seq2_index == len2) {
424 <                diag_lower = k + 1;
424 >                diag_lower = k + 1;
425                  end2_reached = true;
426              }
427              if (seq1_index == len1) {
428 <                diag_upper = k - 1;
428 >                diag_upper = k - 1;
429                  end1_reached = true;
430              }
431          }  // end loop over diagonals
432  
433          // compute the maximum score possible for distance d
434 <        curr_score = curr_extent * (match_cost / 2) -
434 >        curr_score = curr_extent * (match_cost / 2) -
435                          d * (match_cost + mismatch_cost);
436          // if this is the best score seen so far, update the
437          // statistics of the best alignment
# Line 417 | Line 441
441              best_diag = curr_diag;
442              seq2_align_len = curr_seq2_index;
443              seq1_align_len = curr_seq2_index + best_diag - diag_origin;
444 <        }
444 >        }
445          else {
446              max_score[d] = max_score[d - 1];
447          }
# Line 428 | Line 452
452          if (diag_lower > diag_upper)
453              break;
454  
455 <        /* set up for the next distance to examine. Because the
456 <           bounds increase by at most one for each distance,
457 <           diag_lower and diag_upper can each be of size at
455 >        /* set up for the next distance to examine. Because the
456 >           bounds increase by at most one for each distance,
457 >           diag_lower and diag_upper can each be of size at
458             most max_diags+2 */
459  
460          if (!end2_reached)
461 <            diag_lower--;
461 >            diag_lower--;
462          if (!end1_reached)
463              diag_upper++;
464  
# Line 455 | Line 479
479              }
480      }   // end loop over distinct distances
481  
482 <    
482 >
483      if (edit_block == NULL)
484          return best_dist;
485  
486      //----  perform traceback
487 <    d = best_dist;
487 >    d = best_dist;
488      seq1_index = seq1_align_len;
489      seq2_index = seq2_align_len;
490      // for all positive distances
# Line 476 | Line 500
500             traceback operation. best_diag starts off with the
501             value computed during the alignment process */
502  
503 <        new_diag = s_GetNextNonAffineTback(last_seq2_off, d,
503 >        new_diag = s_GetNextNonAffineTback(last_seq2_off, d,
504                                             best_diag, &new_seq2_index);
505  
506          if (new_diag == best_diag) {
# Line 484 | Line 508
508              if (seq2_index - new_seq2_index > 0) {
509                    edit_block->opRep(seq2_index - new_seq2_index);
510              }
511 <        }
511 >        }
512          else if (new_diag < best_diag) {
513              // smaller diagonal: issue a group of substitutions
514              //   and then a gap in seq2 */
# Line 493 | Line 517
517              }
518              //GapPrelimEditBlockAdd(edit_block, eGapAlignIns, 1);
519              edit_block->opIns(1);
520 <        }
520 >        }
521          else {
522              // larger diagonal: issue a group of substitutions
523              //   and then a gap in seq1
# Line 502 | Line 526
526              }
527              edit_block->opDel(1);
528          }
529 <        d--;
530 <        best_diag = new_diag;
531 <        seq2_index = new_seq2_index;
529 >        d--;
530 >        best_diag = new_diag;
531 >        seq2_index = new_seq2_index;
532      }
533   //done:
534      // handle the final group of substitutions back to distance zero,
# Line 526 | Line 550
550        unsigned char op_type = 3 - ( ed_script->ops[i] & gxEDIT_OP_MASK );
551        if (op_type == 3)
552           GError("Error: printEditScript encountered op_type 3 ?!\n");
553 <      GMessage("%d%c ", num, xgapcodes[op_type]);
553 >      GMessage("%d%c ", num, xgapcodes[op_type]);
554        }
555      GMessage("\n");
556    }
# Line 677 | Line 701
701   }
702  
703  
704 < GXBandSet* collectSeeds_R(GList<GXSeed>& seeds, const char* seqa, int a_len, const char* seqb, int b_len) {
704 > GXBandSet* collectSeeds_R(GList<GXSeed>& seeds, const char* seqa, int a_len,
705 >        GVec<uint16>* amers[], const char* seqb, int b_len) {
706   //overlap of right (3') end of seqb
707   //hash the first 12 bases of seqa:
708   int aimin=0;
# Line 738 | Line 763
763   return diagstrips;
764   }
765  
766 < GXBandSet* collectSeeds_L(GList<GXSeed>& seeds, const char* seqa, int a_len, const char* seqb, int b_len) {
766 > GXBandSet* collectSeeds_L(GList<GXSeed>& seeds, const char* seqa, int a_len,
767 >        GVec<uint16>* amers[], const char* seqb, int b_len) {
768   //overlap of left (5') end of seqb
769   //hash the last 12 bases of seqa:
770   int aimin=GMAX(0,(a_len-16));
# Line 888 | Line 914
914                              reward, penalty, s_ext_r, q_ext_r, *gxmem, ed_script_fwd);
915      //check extension here and bail out if not a good right extension
916      //assuming s_max is really at the right end of s_seq
917 <    if (trim!=NULL && trim->type==galn_TrimRight && s_alnstart+s_ext_r<trim->boundary) {
917 >    if (trim!=NULL && trim->type==galn_TrimRight &&
918 >           s_alnstart+s_ext_r<trim->boundary) {
919        delete ed_script_fwd;
920        if (freeAlnMem) delete gxmem;
921        return NULL;
# Line 943 | Line 970
970       reward, penalty, xdrop, gxmem, trim, editscript);
971   }
972  
973 < GXAlnInfo* match_RightEnd(const char* seqa, int seqa_len, const char* seqb, int seqb_len,
974 <            CGreedyAlignData* gxmem, int min_pid) {
973 > GXAlnInfo* match_RightEnd(const char* seqa, int seqa_len, GVec<uint16>* amers[],
974 >                 const char* seqb, int seqb_len, CGreedyAlignData* gxmem, int min_pid) {
975    bool editscript=false;
976    #ifdef GDEBUG
977     editscript=true;
# Line 953 | Line 980
980  
981    CAlnTrim trimInfo(galn_TrimRight, seqb, seqb_len);
982    GList<GXSeed> rseeds(true,true,false);
983 <  GXBandSet* alnbands=collectSeeds_R(rseeds, seqa, seqa_len, seqb, seqb_len);
983 >  GXBandSet* alnbands=collectSeeds_R(rseeds, seqa, seqa_len, amers, seqb, seqb_len);
984    GList<GXSeed> anchor_seeds(cmpSeedDiag, NULL, true); //stores unique seeds per diagonal
985    //did we find a shortcut?
986    if (alnbands->qmatch) {
# Line 1050 | Line 1077
1077    else return NULL;
1078   }
1079  
1080 < GXAlnInfo* match_LeftEnd(const char* seqa, int seqa_len, const char* seqb, int seqb_len,
1081 <          CGreedyAlignData* gxmem, int min_pid) {
1080 > GXAlnInfo* match_LeftEnd(const char* seqa, int seqa_len, GVec<uint16>* amers[],
1081 >                 const char* seqb, int seqb_len, CGreedyAlignData* gxmem, int min_pid) {
1082    bool editscript=false;
1083    #ifdef GDEBUG
1084     editscript=true;
# Line 1059 | Line 1086
1086    #endif
1087    CAlnTrim trimInfo(galn_TrimLeft, seqb, seqb_len);
1088    GList<GXSeed> rseeds(true,true,false);
1089 <  GXBandSet* alnbands=collectSeeds_L(rseeds, seqa, seqa_len, seqb, seqb_len);
1089 >  GXBandSet* alnbands=collectSeeds_L(rseeds, seqa, seqa_len, amers, seqb, seqb_len);
1090    GList<GXAlnInfo> galns(true,true,false);
1091    GList<GXSeed> anchor_seeds(cmpSeedDiag, NULL, true); //stores unique seeds per diagonal
1092  

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines