void RefineW(const MSA &msaIn, MSA &msaOut) { const unsigned uSeqCount = msaIn.GetSeqCount(); const unsigned uColCount = msaIn.GetColCount(); // Reserve same nr seqs, 20% more cols const unsigned uReserveColCount = (uColCount*120)/100; msaOut.SetSize(uSeqCount, uReserveColCount); for (unsigned uSeqIndex = 0; uSeqIndex < uSeqCount; ++uSeqIndex) { msaOut.SetSeqName(uSeqIndex, msaIn.GetSeqName(uSeqIndex)); msaOut.SetSeqId(uSeqIndex, msaIn.GetSeqId(uSeqIndex)); } const unsigned uWindowCount = (uColCount + g_uRefineWindow.get() - 1)/g_uRefineWindow.get(); if (0 == g_uWindowTo.get()) g_uWindowTo.get() = uWindowCount - 1; #if MEMDEBUG _CrtSetBreakAlloc(1560); #endif if (g_uWindowOffset.get() > 0) { MSA msaTmp; MSAFromColRange(msaIn, 0, g_uWindowOffset.get(), msaOut); } if (!g_bQuiet.get()) fprintf(stderr, "\n"); for (unsigned uWindowIndex = g_uWindowFrom.get(); uWindowIndex <= g_uWindowTo.get(); ++uWindowIndex) { if (!g_bQuiet.get()) fprintf(stderr, "Window %d of %d \r", uWindowIndex, uWindowCount); const unsigned uColFrom = g_uWindowOffset.get() + uWindowIndex*g_uRefineWindow.get(); unsigned uColTo = uColFrom + g_uRefineWindow.get() - 1; if (uColTo >= uColCount) uColTo = uColCount - 1; assert(uColTo >= uColFrom); SeqVect v; SeqVectFromMSACols(msaIn, uColFrom, uColTo, v); #if MEMDEBUG _CrtMemState s1; _CrtMemCheckpoint(&s1); #endif // Begin AED 5/20/06 // remove any empty seqs in this window std::vector< size_t > empty_seqs; SeqVect vr; for( size_t seqI = 0; seqI < v.size(); ++seqI ) { if( v[seqI]->size() == 0 ) empty_seqs.push_back(seqI); else vr.push_back(v[seqI]); } std::vector< unsigned > seqid_map( vr.size() ); for( size_t seqI = 0; seqI < vr.size(); ++seqI ) { seqid_map[seqI] = vr[seqI]->GetId(); vr[seqI]->SetId(seqI); } MSA msaTmp; if( vr.size() > 1 ) MUSCLE(vr, msaTmp); // remap the seqids to their original state for( size_t seqI = 0; seqI < vr.size(); ++seqI ) vr[seqI]->SetId(seqid_map[seqI]); // merge empty seqs back in { const unsigned uSeqCount = msaOut.GetSeqCount(); const unsigned uColCount1 = msaOut.GetColCount(); const unsigned uColCount2 = vr.size() > 1 ? msaTmp.GetColCount() : vr[0]->size(); const unsigned uColCountCat = uColCount1 + uColCount2; for( unsigned seqI = 0; seqI < vr.size(); ++seqI ) { unsigned uSeqIndex = msaOut.GetSeqIndex(seqid_map[seqI]); if( vr.size() > 1 ) { unsigned uSeqIndex2 = msaTmp.GetSeqIndex(seqI); for (unsigned uColIndex = 0; uColIndex < uColCount2; ++uColIndex) { const char c = msaTmp.GetChar(uSeqIndex2, uColIndex); msaOut.SetChar(uSeqIndex, uColCount1 + uColIndex, c); } }else{ for (unsigned uColIndex = 0; uColIndex < uColCount2; ++uColIndex) { const char c = vr[0]->GetChar(uColIndex); msaOut.SetChar(uSeqIndex, uColCount1 + uColIndex, c); } } } for( unsigned seqI = 0; seqI < empty_seqs.size(); ++seqI ) { unsigned uSeqId2 = v[empty_seqs[seqI]]->GetId(); unsigned uSeqIndex = msaOut.GetSeqIndex(uSeqId2); for (unsigned uColIndex = 0; uColIndex < uColCount2; ++uColIndex) { msaOut.SetChar(uSeqIndex, uColCount1 + uColIndex, '-'); } } vr.clear(); } // AppendMSA(msaOut, msaTmp); // end AED 5/20/06 if (uWindowIndex == g_uSaveWindow.get()) { MSA msaInTmp; unsigned uOutCols = msaOut.GetColCount(); unsigned un = uColTo - uColFrom + 1; MSAFromColRange(msaIn, uColFrom, un, msaInTmp); char fn[256]; sprintf(fn, "win%d_inaln.tmp", uWindowIndex); TextFile fIn(fn, true); msaInTmp.ToFile(fIn); sprintf(fn, "win%d_inseqs.tmp", uWindowIndex); TextFile fv(fn, true); v.ToFile(fv); sprintf(fn, "win%d_outaln.tmp", uWindowIndex); TextFile fOut(fn, true); msaTmp.ToFile(fOut); } #if MEMDEBUG void FreeDPMemSPN(); FreeDPMemSPN(); _CrtMemState s2; _CrtMemCheckpoint(&s2); _CrtMemState s; _CrtMemDifference(&s, &s1, &s2); _CrtMemDumpStatistics(&s); _CrtMemDumpAllObjectsSince(&s1); exit(1); #endif //#if DEBUG // AssertMSAEqIgnoreCaseAndGaps(msaInTmp, msaTmp); //#endif } if (!g_bQuiet.get()) fprintf(stderr, "\n"); // AssertMSAEqIgnoreCaseAndGaps(msaIn, msaOut);//@@uncomment! }
static SCORE ScoreSeqPair(const MSA &msa1, unsigned uSeqIndex1, const MSA &msa2, unsigned uSeqIndex2, SCORE *ptrLetters, SCORE *ptrGaps) { g_ptrMSA1.get() = &msa1; g_ptrMSA2.get() = &msa2; g_uSeqIndex1.get() = uSeqIndex1; g_uSeqIndex2.get() = uSeqIndex2; const unsigned uColCount = msa1.GetColCount(); const unsigned uColCount2 = msa2.GetColCount(); if (uColCount != uColCount2) Quit("ScoreSeqPair, different lengths"); #if TRACE Log("ScoreSeqPair\n"); Log("%16.16s ", msa1.GetSeqName(uSeqIndex1)); for (unsigned i = 0; i < uColCount; ++i) Log("%c", msa1.GetChar(uSeqIndex1, i)); Log("\n"); Log("%16.16s ", msa2.GetSeqName(uSeqIndex2)); for (unsigned i = 0; i < uColCount; ++i) Log("%c", msa1.GetChar(uSeqIndex2, i)); Log("\n"); #endif SCORE scoreTotal = 0; // Substitution scores unsigned uFirstLetter1 = uInsane; unsigned uFirstLetter2 = uInsane; unsigned uLastLetter1 = uInsane; unsigned uLastLetter2 = uInsane; for (unsigned uColIndex = 0; uColIndex < uColCount; ++uColIndex) { bool bGap1 = msa1.IsGap(uSeqIndex1, uColIndex); bool bGap2 = msa2.IsGap(uSeqIndex2, uColIndex); bool bWildcard1 = msa1.IsWildcard(uSeqIndex1, uColIndex); bool bWildcard2 = msa2.IsWildcard(uSeqIndex2, uColIndex); if (!bGap1) { if (uInsane == uFirstLetter1) uFirstLetter1 = uColIndex; uLastLetter1 = uColIndex; } if (!bGap2) { if (uInsane == uFirstLetter2) uFirstLetter2 = uColIndex; uLastLetter2 = uColIndex; } if (bGap1 || bGap2 || bWildcard1 || bWildcard2) continue; unsigned uLetter1 = msa1.GetLetter(uSeqIndex1, uColIndex); unsigned uLetter2 = msa2.GetLetter(uSeqIndex2, uColIndex); SCORE scoreMatch = (*g_ptrScoreMatrix.get())[uLetter1][uLetter2]; scoreTotal += scoreMatch; #if TRACE Log("%c <-> %c = %7.1f %10.1f\n", msa1.GetChar(uSeqIndex1, uColIndex), msa2.GetChar(uSeqIndex2, uColIndex), scoreMatch, scoreTotal); #endif } *ptrLetters = scoreTotal; // Gap penalties unsigned uGapLength = uInsane; unsigned uGapStartCol = uInsane; bool bGapping1 = false; bool bGapping2 = false; for (unsigned uColIndex = 0; uColIndex < uColCount; ++uColIndex) { bool bGap1 = msa1.IsGap(uSeqIndex1, uColIndex); bool bGap2 = msa2.IsGap(uSeqIndex2, uColIndex); if (bGap1 && bGap2) continue; if (bGapping1) { if (bGap1) ++uGapLength; else { bGapping1 = false; bool bNTerm = (uFirstLetter2 == uGapStartCol); bool bCTerm = (uLastLetter2 + 1 == uColIndex); SCORE scoreGap = GapPenalty(uGapLength, bNTerm || bCTerm); scoreTotal += scoreGap; #if TRACE LogGap(uGapStartCol, uColIndex - 1, uGapLength, bNTerm, bCTerm); Log("GAP %7.1f %10.1f\n", scoreGap, scoreTotal); #endif } continue; } else { if (bGap1) { uGapStartCol = uColIndex; bGapping1 = true; uGapLength = 1; continue; } } if (bGapping2) { if (bGap2) ++uGapLength; else { bGapping2 = false; bool bNTerm = (uFirstLetter1 == uGapStartCol); bool bCTerm = (uLastLetter1 + 1 == uColIndex); SCORE scoreGap = GapPenalty(uGapLength, bNTerm || bCTerm); scoreTotal += scoreGap; #if TRACE LogGap(uGapStartCol, uColIndex - 1, uGapLength, bNTerm, bCTerm); Log("GAP %7.1f %10.1f\n", scoreGap, scoreTotal); #endif } } else { if (bGap2) { uGapStartCol = uColIndex; bGapping2 = true; uGapLength = 1; } } } if (bGapping1 || bGapping2) { SCORE scoreGap = GapPenalty(uGapLength, true); scoreTotal += scoreGap; #if TRACE LogGap(uGapStartCol, uColCount - 1, uGapLength, false, true); Log("GAP %7.1f %10.1f\n", scoreGap, scoreTotal); #endif } *ptrGaps = scoreTotal - *ptrLetters; return scoreTotal; }
SCORE ScoreGaps(const MSA &msa, const unsigned DiffCols[], unsigned DiffColCount) { MuscleContext *ctx = getMuscleContext(); unsigned &g_ColCount = ctx->scoregaps.g_ColCount; unsigned &g_MaxSeqCount = ctx->scoregaps.g_MaxSeqCount; unsigned &g_MaxColCount = ctx->scoregaps.g_MaxColCount; GAPINFO** &g_Gaps = ctx->scoregaps.g_Gaps; bool* &g_ColDiff = ctx->scoregaps.g_ColDiff; #if TRACE { Log("ScoreGaps\n"); Log("DiffCols "); for (unsigned i = 0; i < DiffColCount; ++i) Log(" %u", DiffCols[i]); Log("\n"); Log("msa=\n"); msa.LogMe(); Log("\n"); } #endif const unsigned SeqCount = msa.GetSeqCount(); const unsigned ColCount = msa.GetColCount(); g_ColCount = ColCount; if (SeqCount > g_MaxSeqCount) { delete[] g_Gaps; g_MaxSeqCount = SeqCount + 256; g_Gaps = new GAPINFO *[g_MaxSeqCount]; } memset(g_Gaps, 0, SeqCount*sizeof(GAPINFO *)); if (ColCount > g_MaxColCount) { delete[] g_ColDiff; g_MaxColCount = ColCount + 256; g_ColDiff = new bool[g_MaxColCount]; } memset(g_ColDiff, 0, g_ColCount*sizeof(bool)); for (unsigned i = 0; i < DiffColCount; ++i) { unsigned Col = DiffCols[i]; assert(Col < ColCount); g_ColDiff[Col] = true; } for (unsigned SeqIndex = 0; SeqIndex < SeqCount; ++SeqIndex) FindIntersectingGaps(msa, SeqIndex); #if TRACE { Log("\n"); Log("Intersecting gaps:\n"); Log(" "); for (unsigned Col = 0; Col < ColCount; ++Col) Log("%c", g_ColDiff[Col] ? '*' : ' '); Log("\n"); Log(" "); for (unsigned Col = 0; Col < ColCount; ++Col) Log("%d", Col%10); Log("\n"); for (unsigned Seq = 0; Seq < SeqCount; ++Seq) { Log("%3d: ", Seq); for (unsigned Col = 0; Col < ColCount; ++Col) Log("%c", msa.GetChar(Seq, Col)); Log(" :: "); for (GAPINFO *GI = g_Gaps[Seq]; GI; GI = GI->Next) Log(" (%d,%d)", GI->Start, GI->End); Log(" >%s\n", msa.GetSeqName(Seq)); } Log("\n"); } #endif SCORE Score = 0; for (unsigned Seq1 = 0; Seq1 < SeqCount; ++Seq1) { const WEIGHT w1 = msa.GetSeqWeight(Seq1); for (unsigned Seq2 = Seq1 + 1; Seq2 < SeqCount; ++Seq2) { const WEIGHT w2 = msa.GetSeqWeight(Seq2); // const SCORE Pair = ScorePair(Seq1, Seq2); const SCORE Pair = ScoreSeqPairGaps(msa, Seq1, msa, Seq2); Score += w1*w2*Pair; #if TRACE Log("Seq1=%u Seq2=%u ScorePair=%.4g w1=%.4g w2=%.4g Sum=%.4g\n", Seq1, Seq2, Pair, w1, w2, Score); #endif } } return Score; }