ViewVC Help
View File | Revision Log | Show Annotations | View Changeset | Root Listing
root/gclib/gclib/GAlnExtend.cpp
(Generate patch)
# Line 205 | Line 205
205  
206   int GXGreedyExtend(const char* seq1, int len1,
207                         const char* seq2, int len2,
208 <                       bool reverse, int xdrop_threshold,
209 <                       int match_cost, int mismatch_cost,
208 >                       bool reverse, //int xdrop_threshold, int match_cost, int mismatch_cost,
209                         int& seq1_align_len, int& seq2_align_len,
210                         CGreedyAlignData& aux_data,
211                         GXEditScript *edit_block) {
# Line 275 | Line 274
274         xdrop_offset gives the distance backwards in the score
275         array to look */
276  
277 <    xdrop_offset = (xdrop_threshold + match_cost / 2) /
279 <                           (match_cost + mismatch_cost) + 1;
277 >    xdrop_offset = aux_data.xdrop_ofs;
278  
279      // find the offset of the first mismatch between seq1 and seq2
280  
# Line 327 | Line 325
325      // fill in the initial offsets of the distance matrix
326  
327      last_seq2_off[0][diag_origin] = seq1_index;
328 <    max_score[0] = seq1_index * match_cost;
328 >    max_score[0] = seq1_index * aux_data.match_reward;
329      diag_lower = diag_origin - 1;
330      diag_upper = diag_origin + 1;
331      end1_reached = end2_reached = false;
# Line 353 | Line 351
351          // compute the score for distance d corresponding to the X-dropoff criterion
352  
353          xdrop_score = max_score[d - xdrop_offset] +
354 <                      (match_cost + mismatch_cost) * d - xdrop_threshold;
355 <        xdrop_score = (int)ceil((double)xdrop_score / (match_cost / 2));
354 >                      (aux_data.match_reward + aux_data.mismatch_penalty) * d - aux_data.x_drop;
355 >        xdrop_score = (int)ceil((double)xdrop_score / (aux_data.match_reward>>1));
356  
357          // for each diagonal of interest
358          for (k = tmp_diag_lower; k <= tmp_diag_upper; k++) {
# Line 431 | Line 429
429          }  // end loop over diagonals
430  
431          // compute the maximum score possible for distance d
432 <        curr_score = curr_extent * (match_cost / 2) -
433 <                        d * (match_cost + mismatch_cost);
432 >        curr_score = curr_extent * (aux_data.match_reward / 2) -
433 >                        d * (aux_data.match_reward + aux_data.mismatch_penalty);
434          // if this is the best score seen so far, update the
435          // statistics of the best alignment
436          if (curr_score > max_score[d - 1]) {
# Line 465 | Line 463
463          if (edit_block == NULL) {
464             // if no traceback is specified, the next row of
465             //   last_seq2_off can reuse previously allocated memory
466 <           //TODO FIXME The following assumes two arrays of
466 >           //WARNING The following assumes two arrays of
467             //  at least max_dist+4 int's have already been allocated
468              last_seq2_off[d + 1] = last_seq2_off[d - 1];
469              }
# Line 474 | Line 472
472              // so a new row must be allocated
473              last_seq2_off[d + 1] = (int*)mem_pool->getByteSpace((diag_upper - diag_lower + 7)*sizeof(int));
474              // move the origin for this row backwards
475 <            //TODO FIXME: dubious pointer arithmetic ?!
475 >            // dubious pointer arithmetic ?!
476              //last_seq2_off[d + 1] = last_seq2_off[d + 1] - diag_lower + 2;
477              }
478      }   // end loop over distinct distances
# Line 604 | Line 602
602   const int a_dropoff_score=7;
603   const int a_min_score=12; //at least 6 bases full match
604  
605 < // ------------------ adapter matching - simple k-mer seed & extend, no indels for now
605 > // ------------------ adaptor matching - simple k-mer seed & extend, no indels for now
606   //when a k-mer match is found, simply try to extend the alignment using a drop-off scheme
607   //check minimum score and
608 < //for 3' adapter trimming:
608 > //for 3' adaptor trimming:
609   //     require that the right end of the alignment for either the adaptor OR the read must be
610   //     < 3 distance from its right end
611 < // for 5' adapter trimming:
611 > // for 5' adaptor trimming:
612   //     require that the left end of the alignment for either the adaptor OR the read must
613   //     be at coordinate < 3 from start
614  
# Line 683 | Line 681
681    int mmovh5=(a_l<b_l)? a_l : b_l;
682    if (maxscore>=a_min_score && mmovh3<2 && mmovh5<2) {
683       if (a_l<a_ovh3) {
684 <        //adapter closer to the left end (typical for 5' adapter)
684 >        //adaptor closer to the left end (typical for 5' adaptor)
685          l5=a_r+1;
686          l3=alen-1;
687          }
688        else {
689 <        //adapter matching at the right end (typical for 3' adapter)
689 >        //adaptor matching at the right end (typical for 3' adaptor)
690          l5=0;
691          l3=a_l-1;
692          }
# Line 700 | Line 698
698    return false;
699   }
700  
701 < /*
704 < GXBandSet* collectSeeds_R(GList<GXSeed>& seeds, GXSeqData& sd) {
705 < int bimin=GMAX(0,(sd.blen-sd.alen-6));
706 < GXSeedTable gx(sd.alen, sd.blen);
707 < GXBandSet* diagstrips=new GXBandSet(sd.alen, sd.blen); //set of overlapping 3-diagonal strips
708 < //for (int bi=0;bi<=b_len-6;bi++) {
709 < for (int bi=sd.blen-6;bi>=0;bi--) {
710 <   //for each 6-mer of seqb
711 <   uint16 bv = get6mer((char*)&(sd.bseq[bi]));
712 <   GVec<uint16>* alocs = sd.amers[bv];
713 <   if (alocs==NULL) continue;
714 <   //extend each hit
715 <   for (int h=0;h<alocs->Count();h++) {
716 <         int ai=alocs->Get(h);
717 <         //word match
718 <         if (gx.x(ai,bi))
719 <           //already have a previous seed covering this region of this diagonal
720 <           continue;
721 <         for (int i=0;i<6;i++)
722 <            gx.x(ai+i,bi+i)=1;
723 <         //see if we can extend to the left
724 <         int aix=ai-1;
725 <         int bix=bi-1;
726 <         int len=6;
727 <         while (bix>=0 && aix>=0 && sd.aseq[aix]==sd.bseq[bix]) {
728 <                 gx.x(aix,bix)=1;
729 <                 ai=aix;
730 <                 bi=bix;
731 <                 aix--;bix--;
732 <                 len++;
733 <                 }
734 <        if (len>sd.amlen) {
735 <                //heuristics: likely the best we can get
736 <                //quick match shortcut
737 <                diagstrips->qmatch=new GXSeed(ai,bi,len);
738 <                return diagstrips;
739 <                }
740 <         if (bi<bimin && len<9) continue; //skip middle seeds that are not high scoring enough
741 <         GXSeed* newseed=new GXSeed(ai,bi,len);
742 <         seeds.Add(newseed);
743 <         diagstrips->addSeed(newseed);//add it to all 3 adjacent diagonals
744 <         //special last resort terminal match to be used if no better alignment is there
745 <         if (ai<2 && bi+len>sd.blen-2 &&
746 <                 (!diagstrips->tmatch || diagstrips->tmatch->len<len)) diagstrips->tmatch=newseed;
747 <         }
748 <  }
749 < for (int i=0;i<diagstrips->Count();i++) {
750 <    diagstrips->Get(i)->finalize(); //adjust scores due to overlaps or gaps between seeds
751 <    }
752 < diagstrips->setSorted(true); //sort by score
753 < return diagstrips;
754 < }
755 < */
756 < GXBandSet* collectSeeds(GList<GXSeed>& seeds, GXSeqData& sd) {
757 < int bimax=GMIN((sd.alen+2), (sd.blen-6));
701 > GXBandSet* collectSeeds(GList<GXSeed>& seeds, GXSeqData& sd, GAlnTrimType trim_intent) {
702   int bimin=GMAX(0,(sd.blen-sd.alen-6)); //from collectSeeds_R
703 + int bimax=GMIN((sd.alen+2), (sd.blen-6));
704 + int b_start = (trim_intent==galn_TrimRight) ? bimin : 0;
705 + int b_end = (trim_intent==galn_TrimLeft) ?  bimax : sd.blen-6;
706   //gx.init(a_maxlen, b_maxlen);
707   GXSeedTable gx(sd.alen, sd.blen);
708   GXBandSet* diagstrips=new GXBandSet(sd.alen, sd.blen); //set of overlapping 3-diagonal strips
709 < for (int bi=0;bi<=sd.blen-6;bi++) {
709 > for (int bi=b_start;bi<=b_end;bi++) {
710     //for each 6-mer of seqb
711     uint16 bv = get6mer((char*) & (sd.bseq[bi]));
712     GVec<uint16>* alocs = sd.amers[bv];
# Line 771 | Line 718
718           if (gx.x(ai,bi))
719             //already have a previous seed covering this region of this diagonal
720             continue;
721 +         if (trim_intent==galn_TrimLeft && sd.blen>sd.alen+6 && bi>ai+6)
722 +               continue; //improper positioning for 5' trimming
723 +         if (trim_intent==galn_TrimRight && sd.blen>sd.alen+6 && bi<ai-6)
724 +               continue; //improper positioning for 5' trimming
725 +
726 +         //TODO: there could be Ns in this seed, should we count/adjust score?
727           for (int i=0;i<6;i++)
728              gx.x(ai+i,bi+i)=1;
729           //see if we can extend to the right
# Line 788 | Line 741
741                  diagstrips->qmatch=new GXSeed(ai,bi,len);
742                  return diagstrips;
743                  }
744 <         if (bi>bimax && len<9) continue; //skip middle seeds that are not high scoring enough
745 <         if (bi<bimin && len<9) continue; //collectSeeds_R
744 >         if (bi>bimax && bi<bimin && len<9)
745 >                 //ignore mid-sequence seeds that are not high scoring
746 >             continue;
747  
748           GXSeed* newseed=new GXSeed(ai,bi,len);
749           seeds.Add(newseed);
750           diagstrips->addSeed(newseed);//add it to all 3 adjacent diagonals
751 <     //special last resort terminal match to be used if no better alignment is there
751 >     //keep last resort terminal match to be used if no better alignment is there
752       if (bi<2 && ai+len>=sd.alen-1 &&
753                   (!diagstrips->tmatch_l || diagstrips->tmatch_l->len<len))
754                        diagstrips->tmatch_l=newseed;
# Line 863 | Line 817
817    s_alnstart--;
818    if (q_seq==NULL || q_seq[0]==0 || s_seq==NULL || s_seq[0]==0)
819      GError("GreedyAlign() Error: attempt to use an empty sequence string!\n");
820 <  if (q_seq[q_alnstart]!=s_seq[s_alnstart])
821 <    GError("GreedyAlign() Error: improper anchor (mismatch)!\n");
820 >  /*if (q_seq[q_alnstart]!=s_seq[s_alnstart])
821 >    GError("GreedyAlign() Error: improper anchor (mismatch):\n%s (start %d len %d)\n%s (start %d len %d)\n",
822 >           q_seq, q_alnstart, q_max, s_seq, s_alnstart, s_max);
823 >           */
824    int q_ext_l=0, q_ext_r=0, s_ext_l=0, s_ext_r=0;
825    const char* q=q_seq+q_alnstart;
826    int q_avail=q_max-q_alnstart;
827    const char* s=s_seq+s_alnstart;
828    int s_avail=s_max-s_alnstart;
829    if (penalty<0) penalty=-penalty;
874  int MIN_GREEDY_SCORE=6*reward; //minimum score for an alignment to be reported for 0 diffs
830    GXAlnInfo* alninfo=NULL;
831    bool freeAlnMem=(gxmem==NULL);
832    if (freeAlnMem) {
833       gxmem=new CGreedyAlignData(reward, penalty, xdrop);
834 +     reward=gxmem->match_reward;
835 +     penalty=gxmem->mismatch_penalty;
836 +     xdrop=gxmem->x_drop;
837       }
838 <  else gxmem->reset();
838 >   else
839 >         gxmem->reset();
840 >  int minMatch= trim ? trim->minMatch : 6;
841 >  int MIN_GREEDY_SCORE=minMatch*reward; //minimum score for an alignment to be reported for 0 diffs
842    int retscore = 0;
843    int numdiffs = 0;
844    if (trim!=NULL && trim->type==galn_TrimLeft) {
# Line 885 | Line 846
846      if (editscript)
847         ed_script_rev=new GXEditScript();
848  
849 <    int numdiffs_l = GXGreedyExtend(s_seq, s_alnstart, q_seq, q_alnstart, true, xdrop,
850 <                   reward, penalty, s_ext_l, q_ext_l, *gxmem, ed_script_rev);
849 >    int numdiffs_l = GXGreedyExtend(s_seq, s_alnstart, q_seq, q_alnstart, true, // xdrop, reward, penalty,
850 >                    s_ext_l, q_ext_l, *gxmem, ed_script_rev);
851      //check this extension here and bail out if it's not a good extension
852      if (s_ext_l+(trim->seedlen>>1) < trim->safelen &&
853 <        q_alnstart+1-q_ext_l>trim->boundary &&
854 <        s_alnstart+1-s_ext_l>trim->boundary) {
853 >        q_alnstart+1-q_ext_l>1 &&
854 >        s_alnstart+1-s_ext_l>trim->l_boundary) {
855        delete ed_script_rev;
856        if (freeAlnMem) delete gxmem;
857        return NULL;
# Line 898 | Line 859
859  
860      if (editscript)
861         ed_script_fwd=new GXEditScript();
862 <    int numdiffs_r = GXGreedyExtend(s, s_avail, q, q_avail, false, xdrop,
863 <                            reward, penalty, s_ext_r, q_ext_r, *gxmem, ed_script_fwd);
862 >    int numdiffs_r = GXGreedyExtend(s, s_avail, q, q_avail, false, //xdrop, reward, penalty,
863 >                                     s_ext_r, q_ext_r, *gxmem, ed_script_fwd);
864      numdiffs=numdiffs_r+numdiffs_l;
865      //convert num diffs to actual score
866 <    retscore = (q_ext_r + s_ext_r + q_ext_l + s_ext_l)*reward/2 - numdiffs*(reward+penalty);
866 >    retscore = (q_ext_r + s_ext_r + q_ext_l + s_ext_l)*(reward>>1) - numdiffs*(reward+penalty);
867      if (editscript)
868         ed_script_rev->Append(ed_script_fwd); //combine the two extensions
869      }
# Line 910 | Line 871
871      if (editscript) {
872         ed_script_fwd=new GXEditScript();
873         }
874 <    int numdiffs_r = GXGreedyExtend(s, s_avail, q, q_avail, false, xdrop,
875 <                            reward, penalty, s_ext_r, q_ext_r, *gxmem, ed_script_fwd);
874 >    int numdiffs_r = GXGreedyExtend(s, s_avail, q, q_avail, false, // xdrop, reward, penalty,
875 >                                     s_ext_r, q_ext_r, *gxmem, ed_script_fwd);
876      //check extension here and bail out if not a good right extension
877      //assuming s_max is really at the right end of s_seq
878      if (trim!=NULL && trim->type==galn_TrimRight &&
879          s_ext_r+(trim->seedlen>>1) < trim->safelen &&
880 <            q_alnstart+q_ext_r<q_max-3 &&
881 <            s_alnstart+s_ext_r<trim->boundary) {
880 >            q_alnstart+q_ext_r<q_max-2 &&
881 >            s_alnstart+s_ext_r<trim->r_boundary) {
882        delete ed_script_fwd;
883        if (freeAlnMem) delete gxmem;
884        return NULL;
885        }
886      if (editscript)
887         ed_script_rev=new GXEditScript();
888 <    int numdiffs_l =  GXGreedyExtend(s_seq, s_alnstart, q_seq, q_alnstart, true, xdrop,
889 <                   reward, penalty, s_ext_l, q_ext_l, *gxmem, ed_script_rev);
888 >    int numdiffs_l =  GXGreedyExtend(s_seq, s_alnstart, q_seq, q_alnstart, true, // xdrop, reward, penalty,
889 >                                  s_ext_l, q_ext_l, *gxmem, ed_script_rev);
890      //convert num diffs to actual score
891      numdiffs=numdiffs_r+numdiffs_l;
892 <    retscore = (q_ext_r + s_ext_r + q_ext_l + s_ext_l)*reward/2 - numdiffs*(reward+penalty);
892 >    retscore = (q_ext_r + s_ext_r + q_ext_l + s_ext_l)*(reward>>1) - numdiffs*(reward+penalty);
893      if (editscript)
894         ed_script_rev->Append(ed_script_fwd); //combine the two extensions
895    }
935
896    if (retscore>=MIN_GREEDY_SCORE) {
897 <    alninfo=new GXAlnInfo(q_seq, q_alnstart+1-q_ext_l,q_alnstart+q_ext_r, s_seq, s_alnstart+1-s_ext_l, s_alnstart+s_ext_r);
897 >    alninfo=new GXAlnInfo(q_seq, q_alnstart+1-q_ext_l, q_alnstart+q_ext_r, s_seq, s_alnstart+1-s_ext_l, s_alnstart+s_ext_r);
898      int hsp_length = GMIN(q_ext_l+q_ext_r, s_ext_l+s_ext_r);
899      alninfo->score=retscore;
900 +    if (gxmem->scaled) alninfo->score >>= 1;
901      alninfo->pid = 100 * (1 - ((double) numdiffs) / hsp_length);
902   #ifdef GDEBUG
903      //if (ed_script_rev) {
# Line 948 | Line 909
909      alninfo->gapinfo = new CAlnGapInfo(ed_script_rev, alninfo->ql-1, alninfo->sl-1);
910      }
911    else {
912 <    if (freeAlnMem) delete gxmem;
912 >    //if (freeAlnMem) delete gxmem;
913      delete ed_script_rev;
914      delete alninfo;
915      alninfo=NULL;
916      }
917 +  if (freeAlnMem) delete gxmem;
918    delete ed_script_fwd;
919    return alninfo;
920   }
# Line 961 | Line 923
923                         const char* s_seq, int s_alnstart, int s_max, CGreedyAlignData* gxmem,
924                         CAlnTrim* trim, bool editscript) {
925   int reward=2;
926 < int penalty=3;
927 < int xdrop=8;
926 > int penalty=10;
927 > int xdrop=32;
928   if (gxmem) {
929     reward=gxmem->match_reward;
930     penalty=gxmem->mismatch_penalty;
# Line 972 | Line 934
934       reward, penalty, xdrop, gxmem, trim, editscript);
935   }
936  
937 < GXAlnInfo* match_RightEnd(GXSeqData& sd, CGreedyAlignData* gxmem, int min_pid) {
937 > GXAlnInfo* match_adaptor(GXSeqData& sd, GAlnTrimType trim_type, int minMatch,
938 >                                     CGreedyAlignData* gxmem, int min_pid) {
939    bool editscript=false;
940    #ifdef GDEBUG
941     editscript=true;
942 <   GMessage("==========> matching Right (3') end : %s\n", sd.aseq);
942 >   if (trim_type==galn_TrimLeft) {
943 >         GMessage("=======> searching left (5') end : %s\n", sd.aseq);
944 >     }
945 >   else if (trim_type==galn_TrimRight) {
946 >     GMessage("=======> searching right(3') end : %s\n", sd.aseq);
947 >     }
948 >   else if (trim_type==galn_TrimEither) {
949 >     GMessage("==========> searching  both ends : %s\n", sd.aseq);
950 >     }
951    #endif
952 <  CAlnTrim trimInfo(galn_TrimRight, sd.bseq, sd.blen, sd.amlen);
952 >  CAlnTrim trimInfo(trim_type, sd.bseq, sd.blen, sd.alen, minMatch, sd.amlen);
953    GList<GXSeed> rseeds(true,true,false);
954 <
984 <  GXBandSet* alnbands=collectSeeds(rseeds, sd);
954 >  GXBandSet* alnbands=collectSeeds(rseeds, sd, trim_type);
955    GList<GXSeed> anchor_seeds(cmpSeedDiag, NULL, true); //stores unique seeds per diagonal
956    //did we find a shortcut?
957    if (alnbands->qmatch) {
# Line 1003 | Line 973
973         GXBand& band=*(alnbands->Get(b));
974         band.seeds.setSorted(cmpSeedScore);
975         anchor_seeds.Add(band.seeds.First());
976 <       band.tested=true;
976 >       //band.tested=true;
977         if (anchor_seeds.Count()>2 || top_band_count>max_top_bands) break;
978         }
979      //#ifdef GDEBUG
# Line 1022 | Line 992
992   #endif
993      GXAlnInfo* alninfo=GreedyAlignRegion(sd.aseq, a1, sd.alen,
994                              sd.bseq, a2, sd.blen, gxmem, &trimInfo, editscript);
995 <    if (alninfo && alninfo->pid>=min_pid &&
1026 <        trimInfo.validate(alninfo->sl, alninfo->sr, alninfo->pid, alninfo->ql-1))
995 >    if (alninfo && alninfo->pid>=min_pid && trimInfo.validate(alninfo))
996               galns.AddIfNew(alninfo, true);
997          else delete alninfo;
998      }
999 <  if (galns.Count()==0 && alnbands->tmatch_r) {
1000 <      //last resort seed
1001 <      GXSeed& aseed=*alnbands->tmatch_r;
1002 <      int halfseed=aseed.len>>1;
1003 <      int a1=aseed.a_ofs+halfseed+1;
1004 <      int a2=aseed.b_ofs+halfseed+1;
1005 <      trimInfo.seedlen=aseed.len;
999 >
1000 >  if (galns.Count()==0) {
1001 >         //last resort: look for weaker terminal seeds
1002 >          GPVec<GXSeed> tmatches(2,false);
1003 >          if (trim_type!=galn_TrimRight) {
1004 >                 if (alnbands->tmatch_l)
1005 >                    tmatches.Add(alnbands->tmatch_l);
1006 >             }
1007 >          if (trim_type!=galn_TrimLeft) {
1008 >                 if (alnbands->tmatch_r)
1009 >                    tmatches.Add(alnbands->tmatch_r);
1010 >             }
1011 >          for (int i=0;i<tmatches.Count();i++) {
1012 >                GXSeed& aseed=*tmatches[i];
1013 >                int halfseed=aseed.len>>1;
1014 >                int a1=aseed.a_ofs+halfseed+1;
1015 >                int a2=aseed.b_ofs+halfseed+1;
1016 >                trimInfo.seedlen=aseed.len;
1017   #ifdef GDEBUG
1018 <    GMessage("\t::: align from seed (%d, %d)of len %d.\n",aseed.a_ofs, aseed.b_ofs,
1018 >    GMessage("\t::: align from terminal seed (%d, %d)of len %d.\n",aseed.a_ofs, aseed.b_ofs,
1019             aseed.len);
1020   #endif
1021 <      GXAlnInfo* alninfo=GreedyAlignRegion(sd.aseq, a1, sd.alen,
1021 >        GXAlnInfo* alninfo=GreedyAlignRegion(sd.aseq, a1, sd.alen,
1022                                  sd.bseq, a2, sd.blen, gxmem, &trimInfo, editscript);
1023 <      if (alninfo && alninfo->pid>=min_pid &&
1044 <           trimInfo.validate(alninfo->sl, alninfo->sr, alninfo->pid, alninfo->ql-1))
1023 >        if (alninfo && alninfo->pid>=min_pid && trimInfo.validate(alninfo))
1024                   galns.AddIfNew(alninfo, true);
1025 <            else delete alninfo;
1025 >             else delete alninfo;
1026 >        }//for each terminal seed
1027        }
1028 <
1049 <  /*
1050 <  //special 3' end case: due to the seed scoring scheme being biased towards the 5' end of the read,
1051 <  //we should also try some seeds closer to the 3' end
1052 <  if (galns.Count()==0) {
1053 <    anchor_seeds.Clear();
1054 <    alnbands->setSorted(cmpDiagBands_R);
1055 <    int max_top_bands=4;
1056 <    int top_band_count=0;
1057 <    //#ifdef GDEBUG
1058 <    //GMessage(":::>> Retrying adjusting sort order.\n");
1059 <    //#endif
1060 <    if (alnbands->tmatch) {
1061 <      //anchor_seeds.setSorted(false);
1062 <      anchor_seeds.Add(alnbands->tmatch);
1063 <      }
1064 <    for (int b=0;b<alnbands->Count();b++) {
1065 <       if (alnbands->Get(b)->score<4) break;
1066 <       //#ifdef GDEBUG
1067 <       //GMessage("\tBand %d score: %d\n", b, alnbands->Get(b)->score);
1068 <       //#endif
1069 <       if (alnbands->Get(b)->tested) continue;
1070 <       top_band_count++;
1071 <       GXBand& band=*(alnbands->Get(b));
1072 <       band.seeds.setSorted(cmpSeedScore);
1073 <       anchor_seeds.Add(band.seeds.First());
1074 <       if (anchor_seeds.Count()>2 || top_band_count>max_top_bands) break;
1075 <       }
1076 <    //#ifdef GDEBUG
1077 <    //GMessage("::: Collected %d anchor seeds.\n",anchor_seeds.Count());
1078 <    //#endif
1079 <    for (int i=0;i<anchor_seeds.Count();i++) {
1080 <      GXSeed& aseed=*anchor_seeds[i];
1081 <      int a1=aseed.a_ofs+(aseed.len>>1)+1;
1082 <      int a2=aseed.b_ofs+(aseed.len>>1)+1;
1083 <      GXAlnInfo* alninfo=GreedyAlignRegion(seqa, a1, seqa_len,
1084 <                              seqb, a2, seqb_len, gxmem, &trimInfo, editscript);
1085 <      if (alninfo && alninfo->pid>=min_pid && trimInfo.validate(alninfo->sl, alninfo->sr, alninfo->pid, alninfo->ql-1))
1086 <               galns.AddIfNew(alninfo, true);
1087 <          else delete alninfo;
1088 <      }
1089 <    } */
1090 <
1091 <  //---- done
1028 >  //---- found all alignments
1029    delete alnbands;
1093  if (galns.Count()) {
1094    GXAlnInfo* bestaln=galns.Shift();
1095    #ifdef GDEBUG
1096      GMessage("Best alignment: a(%d..%d) align to b(%d..%d), score=%d, pid=%4.2f\n", bestaln->ql, bestaln->qr,
1097          bestaln->sl, bestaln->sr, bestaln->score, bestaln->pid);
1098      if (bestaln->gapinfo!=NULL) {
1099        bestaln->gapinfo->printAlignment(stderr, sd.aseq, sd.alen, sd.bseq, sd.blen);
1100        }
1101    #endif
1102
1103    return bestaln;
1104    }
1105  else return NULL;
1106 }
1107
1108 GXAlnInfo* match_LeftEnd(GXSeqData& sd, CGreedyAlignData* gxmem, int min_pid) {
1109  bool editscript=false;
1110  #ifdef GDEBUG
1111   editscript=true;
1112   GMessage("==========> matching Left (5') end : %s\n", sd.aseq);
1113  #endif
1114  CAlnTrim trimInfo(galn_TrimLeft, sd.bseq, sd.blen, sd.amlen);
1115  GList<GXSeed> rseeds(true,true,false);
1116  GXBandSet* alnbands = collectSeeds(rseeds, sd);
1117  GList<GXSeed> anchor_seeds(cmpSeedDiag, NULL, true); //stores unique seeds per diagonal
1118  if (alnbands->qmatch) {
1119    #ifdef GDEBUG
1120     GMessage("::: Found a quick long match at %d, len %d\n",
1121          alnbands->qmatch->b_ofs, alnbands->qmatch->len);
1122    #endif
1123    anchor_seeds.Add(alnbands->qmatch);
1124    }
1125  else {
1126    int max_top_bands=5;
1127    int top_band_count=0;
1128    for (int b=0;b<alnbands->Count();b++) {
1129       if (alnbands->Get(b)->score<6) break;
1130       //#ifdef GDEBUG
1131       //GMessage("\tBand %d score: %d\n", b, alnbands->Get(b)->score);
1132       //#endif
1133       top_band_count++;
1134       GXBand& band=*(alnbands->Get(b));
1135       band.seeds.setSorted(cmpSeedScore);
1136       anchor_seeds.Add(band.seeds.First());
1137       band.tested=true;
1138       if (anchor_seeds.Count()>2 || top_band_count>max_top_bands) break;
1139       }
1140    //#ifdef GDEBUG
1141    //GMessage("::: Collected %d anchor seeds.\n",anchor_seeds.Count());
1142    //#endif
1143    }
1144 GList<GXAlnInfo> galns(true,true,false);
1145 for (int i=0;i<anchor_seeds.Count();i++) {
1146    GXSeed& aseed=*anchor_seeds[i];
1147    int a1=aseed.a_ofs+(aseed.len>>1)+1;
1148    int a2=aseed.b_ofs+(aseed.len>>1)+1;
1149    trimInfo.seedlen=aseed.len;
1150 #ifdef GDEBUG
1151    GMessage("\t::: align from seed (%d, %d)of len %d.\n",aseed.a_ofs, aseed.b_ofs,
1152           aseed.len);
1153 #endif
1154    GXAlnInfo* alninfo=GreedyAlignRegion(sd.aseq, a1, sd.alen,
1155                            sd.bseq, a2, sd.blen, gxmem, &trimInfo, editscript);
1156    if (alninfo && alninfo->pid>=min_pid
1157           && trimInfo.validate(alninfo->sl, alninfo->sr,
1158                    alninfo->pid, sd.alen-alninfo->qr))
1159            galns.AddIfNew(alninfo, true);
1160       else delete alninfo;
1161    }
1162  if (galns.Count()==0 && alnbands->tmatch_l) {
1163      //last resort seed
1164      GXSeed& aseed=*alnbands->tmatch_l;
1165      int a1=aseed.a_ofs+(aseed.len>>1)+1;
1166      int a2=aseed.b_ofs+(aseed.len>>1)+1;
1167      trimInfo.seedlen=aseed.len;
1168      GXAlnInfo* alninfo=GreedyAlignRegion(sd.aseq, a1, sd.alen,
1169                              sd.bseq, a2, sd.blen, gxmem, &trimInfo, editscript);
1170      if (alninfo && alninfo->pid>=min_pid &&
1171        trimInfo.validate(alninfo->sl, alninfo->sr, alninfo->pid, sd.alen-alninfo->qr))
1172         galns.Add(alninfo);
1173      }
1030    /*
1031    #ifdef GDEBUG
1032 <  //print valid alignments found
1032 >  //print all valid alignments found
1033    for (int i=0;i<galns.Count();i++) {
1034      GXAlnInfo* alninfo=galns[i];
1035      GMessage("a(%d..%d) align to b(%d..%d), score=%d, pid=%4.2f\n", alninfo->ql, alninfo->qr,
# Line 1185 | Line 1041
1041      }
1042    #endif
1043    */
1188  //---- done
1189  delete alnbands;
1044    if (galns.Count()) {
1045      GXAlnInfo* bestaln=galns.Shift();
1046      #ifdef GDEBUG
1047        GMessage("Best alignment: a(%d..%d) align to b(%d..%d), score=%d, pid=%4.2f\n", bestaln->ql, bestaln->qr,
1048            bestaln->sl, bestaln->sr, bestaln->score, bestaln->pid);
1049        if (bestaln->gapinfo!=NULL) {
1050 <        bestaln->gapinfo->printAlignment(stderr, sd.aseq, sd.alen,
1197 <                                          sd.bseq, sd.blen);
1050 >        bestaln->gapinfo->printAlignment(stderr, sd.aseq, sd.alen, sd.bseq, sd.blen);
1051          }
1052      #endif
1053      return bestaln;

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines