void DindelUtil::doMultipleReadHaplotypeAlignment(const std::vector<DindelRead> & dReads,
const StringVector & haplotypes)
{
// globally align haplotypes to the first haplotype (arbitrary)
assert(haplotypes.size()>0);
for (size_t h = 0; h < haplotypes.size(); ++h)
{
std::cout << "ALIGNING EVERYTHING AGAINST HAPLOTYPE " << h << "\n";
MultipleAlignment ma;
const std::string rootSequence = haplotypes[h];
ma.addBaseSequence("root", rootSequence, "");
std::string hid;
for(size_t j = 0; j < haplotypes.size(); j++)
{
std::stringstream ss;
if (j!=h)
ss << "haplotype-" << j;
else
ss << "HAPLOTYPE-" << j;
SequenceOverlap o = Overlapper::computeOverlap(rootSequence, haplotypes[j]);
ma.addOverlap(ss.str(), haplotypes[j], "", o);
}
for(size_t r = 0; r < dReads.size(); ++r)
{
std::stringstream ss;
if (r<dReads.size()/2)
ss << "read-" << r << "(" << dReads[r].getID() << ")";
else
ss << "MATE read-" << r;
SequenceOverlap o = Overlapper::computeOverlap(rootSequence, dReads[r].getSequence());
ma.addOverlap(ss.str(), dReads[r].getSequence(), "", o);
}
ma.print(100000);
}
}
bool OverlapHaplotypeBuilder::buildInitialGraph(const StringVector& reads)
{
PROFILE_FUNC("OverlapHaplotypeBuilder::buildInitialGraph")
// Compute initial ordering of reads based on the position of the
// starting kmer sequence. If the starting kmer was corrected out
// of a read, it is discarded.
StringVector ordered_reads;
orderReadsInitial(m_initial_kmer_string, reads, &ordered_reads);
if(ordered_reads.size() < m_parameters.minDiscoveryCount)
return false;
#ifdef SHOW_MULTIPLE_ALIGNMENT
//DEBUG print MA
MultipleAlignment ma = buildMultipleAlignment(ordered_reads);
ma.print(200);
#endif
// Insert initial reads into graph
for(size_t i = 0; i < ordered_reads.size(); ++i)
insertVertexIntoGraph("seed-", ordered_reads[i]);
return true;
}
int IntersectByMaster(CCdCore* ccd, double rowFraction) {
int result = -1;
unsigned int masterLen = (ccd) ? ccd->GetSequenceStringByRow(0).length() : 0;
if (masterLen == 0) return result;
int slaveStart;
int nAlignedIBM = 0;
unsigned int i, j, nBlocks;
unsigned int nRows = ccd->GetNumRows();
// If there is already a consistent block model, do nothing.
MultipleAlignment* ma = new MultipleAlignment(ccd);
if (ma && ma->isBlockAligned()) {
delete ma;
return 0;
}
delete ma;
BlockIntersector blockIntersector(masterLen);
BlockModel* intersectedBlockModel;
//BlockModel* simpleIntersectedBlockModel;
BlockModelPair* bmp;
vector<BlockModelPair*> blockModelPairs;
set<int> forcedCTerminiInIntersection;
list< CRef< CSeq_align > >& cdSeqAligns = ccd->GetSeqAligns();
list< CRef< CSeq_align > >::iterator cdSeqAlignIt = cdSeqAligns.begin(), cdSeqAlignEnd = cdSeqAligns.end();
for (i = 0; cdSeqAlignIt != cdSeqAlignEnd; ++cdSeqAlignIt, ++i) {
bmp = new BlockModelPair(*cdSeqAlignIt);
// We assume # of blocks and all block lengths are same on master and slave.
if (bmp && bmp->isValid()) {
blockModelPairs.push_back(bmp);
blockIntersector.addOneAlignment(bmp->getMaster());
// Find places the intersection can't merge blocks (i.e., where there are
// gaps in the slave across a block boundary, but not in the master).
BlockModel& slave = bmp->getSlave();
nBlocks = slave.getBlocks().size();
for (j = 0; j < nBlocks - 1; ++j) { // '-1' as I don't care about end of the C-terminal block
if (slave.getGapToCTerminal(j) > 0 && bmp->getMaster().getGapToCTerminal(j) == 0) {
forcedCTerminiInIntersection.insert(bmp->getMaster().getBlock(j).getEnd());
}
}
}
}
// There was a problem creating one of the BlockModelPair objects from a seq_align,
// or one or more seq_align was invalid.
if (blockModelPairs.size() != cdSeqAligns.size()) {
return result;
}
//simpleIntersectedBlockModel = blockIntersector.getIntersectedAlignment(forcedCTerminiInIntersection);
intersectedBlockModel = blockIntersector.getIntersectedAlignment(forcedCTerminiInIntersection, rowFraction);
nAlignedIBM = (intersectedBlockModel) ? intersectedBlockModel->getTotalBlockLength() : 0;
if (nAlignedIBM == 0) {
return result;
}
/*
string testStr, testStr2;
string sint = intersectedBlockModel->toString();
string sintsimple = simpleIntersectedBlockModel->toString();
delete simpleIntersectedBlockModel;
cout << "rowFraction = 1:\n" << sintsimple << endl;
cout << "rowFraction = " << rowFraction << ":\n" << sint << endl;
*/
// As we have case where every block model isn't identical,
// change each seq-align to reflect the common set of aligned columns.
nBlocks = intersectedBlockModel->getBlocks().size();
for (i = 0, cdSeqAlignIt = cdSeqAligns.begin(); i < nRows - 1 ; ++i, ++cdSeqAlignIt) {
bmp = blockModelPairs[i]; //BlockModelPair seqAlignPair(*cdSeqAlignIt);
BlockModel* intersectedSeqAlignSlave = new BlockModel(bmp->getSlave().getSeqId(), false);
bmp->reverse();
for (j = 0; j < nBlocks; ++j) {
const Block& jthMasterBlock = intersectedBlockModel->getBlock(j);
slaveStart = bmp->mapToMaster(jthMasterBlock.getStart());
// since we're dealing w/ an intersection, slaveStart should always be valid
assert(slaveStart != -1);
Block b(slaveStart, jthMasterBlock.getLen(), jthMasterBlock.getId());
intersectedSeqAlignSlave->addBlock(b);
}
*cdSeqAlignIt = intersectedSeqAlignSlave->toSeqAlign(*intersectedBlockModel);
//testStr = intersectedSeqAlignSlave->toString();
//testStr2 = bmp->getMaster().toString(); // original *slave* alignment
delete bmp;
}
blockModelPairs.clear();
result = nBlocks;
delete intersectedBlockModel;
return result;
}
void GetAlignmentColumnsForCD(CCdCore* cd, map<unsigned int, string>& columns, unsigned int referenceRow)
{
bool isOK = true, useRefRow = true;
int j;
unsigned int i, col, row, pos, mapIndex, nRows, nCols, nBlocks;
char** alignedResidues = NULL;
string rowString, colString;
// Map column number to position on the selected reference row.
map<unsigned int, unsigned int> colToPos;
map<unsigned int, string> rowStrings;
vector<int> starts, lengths;
CRef< CSeq_align > seqAlign;
// Empty the columns map first, as this is used as a way to flag problems.
columns.clear();
if (!cd) return;
// Check if the block structure is consistent.
try {
MultipleAlignment* ma = new MultipleAlignment(cd);
if (!ma) {
ERR_POST("Creation of MultipleAlignment object failed for CD " << cd->GetAccession() << ".");
return;
} else if (! ma->isBlockAligned()) {
delete ma;
ERR_POST("CD " << cd->GetAccession() << " must have a consistent block structure for column extraction.");
return;
}
delete ma;
ma = NULL;
} catch (...) {
ERR_POST("Could not extract columns for CD " << cd->GetAccession());
}
nCols = cd->GetAlignmentLength();
nRows = cd->GetNumRows();
// Get a reference seq-align for mapping between alignment rows.
// If the columns map index will simply be the column count, use the master, row 0.
if (referenceRow >= nRows) {
useRefRow = false;
referenceRow = 0;
}
if (! cd->GetSeqAlign(referenceRow, seqAlign)) {
isOK = false;
}
// Initialize the column # -> reference row position mapping.
// If useRefRow is true, use the indicated row's coordinates as the position.
// Otherwise, use the column number as the position.
if (isOK && GetBlockStarts(seqAlign, starts, (referenceRow == 0)) > 0 && GetBlockLengths(seqAlign, lengths) > 0) {
nBlocks = starts.size();
if (nBlocks == lengths.size()) {
for (i = 0, col = 0; i < nBlocks; ++i) {
pos = (useRefRow) ? starts[i] : col;
for (j = 0; j < lengths[i]; ++j, ++col, ++pos) {
// Not explicitly checking if 'pos' is aligned since above
// we confirmed the CD has a valid block model.
colToPos[col] = pos;
}
}
} else {
isOK = false;
}
} else {
isOK = false;
}
SetAlignedResiduesForCD(cd, alignedResidues, true);
// Construct the columns as string objects.
if (isOK && alignedResidues) {
for (col = 0; col < nCols; ++col) {
colString.erase();
for (row = 0; row < nRows; ++row) {
colString += alignedResidues[row][col];
}
mapIndex = colToPos[col];
columns[mapIndex] = colString;
}
}
// Clean up array of characters.
if (alignedResidues) {
for (row = 0; row < nRows; ++row) {
delete [] alignedResidues[row];
}
delete [] alignedResidues;
}
}
int main (int argc, char *argv[])
{
int i;
PlatformSupport* Plat = new PlatformSupport();
ColumnComp* CC;
Alignment* ALIGN;
Tree* T;
MultipleAlignment* MA;
ProteinDomains* PROTS =NULL;
MultiAlignRec* pssmAlignment;
char outFileName[STR_LEN];
strcpy(outFileName, "out");
bool colChosen=false, alignChosen=false, treeChosen=false, maChosen=false, usingDomains=false, inputProvided=false, scoresProvided=false;
bool neuralTree=false; bool testing=false;bool testingAcc=false; bool testingTree=false; bool famNames=false; bool treeClusts=false; bool printTreeClusts=false;
bool ma_off=false;
bool tree_loocv=false;//true;
bool silent=false, htmlOutput=false; bool simMatching=false;
bool weighting_on=false;
int matchTopX = TOP_MATCH;
char inputTFs[STR_LEN];
char matchTFs[STR_LEN];
char scoreDist[STR_LEN];
char inputProteins[STR_LEN];
//Misc option settings
bool genRandMotifs=false;
bool genRandScores=false;
char randMatOut[STR_LEN];
char scoresOut[STR_LEN];
//Default alignment settings
double gapOpen = DFLT_GAP_OPEN;
double gapExtend = DFLT_GAP_EXTEND;
bool overlapAlign = DFLT_OVLP_ALIGN;
bool extendOverlap=false;
bool FBP_on = false;
bool preAlign=false;
bool pairwiseOnly=false;
bool forwardAlignOnly=false;
bool ungapped=false;
for(i=1; i<argc; i++){
if(strcmp(argv[i], "-silent")==0)
silent=true;
if(strcmp(argv[i], "-html")==0)
htmlOutput=true;
}
//Welcome message
if(!silent && !htmlOutput){printf("\n\tSTAMP\n\tSimilarity, Tree-building, & Alignment of Motifs and Profiles\n\n\tShaun Mahony\n\tDepartment of Computational Biology\n\tUniversity of Pittsburgh\n\tVersion 1.0 (Winter 2006)\n\n");}
if(argc ==1) //First and Foremost, the help option
{ DisplayHelp();
}else{
for(i=1; i<argc; i++)
{
if(strcmp(argv[i], "-h")==0 || strcmp(argv[i], "?")==0) //First and Foremost, the help option
{ DisplayHelp();
}
if(strcmp(argv[i], "-out")==0) //Output file (for trees & similarity matching)
{ if(argv[i+1]!=NULL)
{ strcpy(outFileName, argv[i+1]);}
}
if(strcmp(argv[i], "-genrand")==0) //Generate random motifs
{ if(argv[i+1]!=NULL)
{ strcpy(randMatOut, argv[i+1]);}
genRandMotifs=true;
}
if(strcmp(argv[i], "-genscores")==0) //Generate simulation scores
{ if(argv[i+1]!=NULL)
{ strcpy(scoresOut, argv[i+1]);}
genRandScores=true;
}
if((strcmp(argv[i], "-cc")) ==0) //Choose a column comparison measure
{
if((strcmp(argv[i+1], "PCC"))==0 || (strcmp(argv[i+1], "pcc"))==0){
CC = new PearsonCorrelation(); //Pearson's correllation coefficient
}else if((strcmp(argv[i+1], "ALLR"))==0 || (strcmp(argv[i+1], "allr"))==0){
CC = new ALLR(); //ALLR
}else if((strcmp(argv[i+1], "ALLR_LL"))==0 || (strcmp(argv[i+1], "allr_ll"))==0){
CC = new ALLR_LL(); //ALLR with lower limit
}else if((strcmp(argv[i+1], "CS"))==0 || (strcmp(argv[i+1], "cs"))==0){
CC = new ChiSq(); //Pearson's Chi Square
}else if((strcmp(argv[i+1], "KL"))==0 || (strcmp(argv[i+1], "kl"))==0){
CC = new KullbackLieber(); //Kullback-Lieber
}else if((strcmp(argv[i+1], "SSD"))==0 || (strcmp(argv[i+1], "ssd"))==0){
CC = new SumSqDiff(); //sum of squared difference
}else{
CC = new PearsonCorrelation(); //Default = PCC
}
colChosen=true;
}
//check for alignment settings
if((strcmp(argv[i], "-go")) ==0){ //Gap Open
if(argv[i+1]!=NULL)
{ gapOpen=strtod(argv[i+1], NULL);}
}
if((strcmp(argv[i], "-ge")) ==0){ //Gap Extend
if(argv[i+1]!=NULL)
{ gapExtend=strtod(argv[i+1], NULL);}
}
if((strcmp(argv[i], "-overlapalign")) ==0){ //Only complete overlapping alignments
overlapAlign = true; if(!silent && !htmlOutput){printf("Overlapping alignments only\n");}
}if((strcmp(argv[i], "-nooverlapalign")) ==0){ //All overlapping alignments
overlapAlign = false;
}
if((strcmp(argv[i], "-extendoverlap")) ==0){
extendOverlap=true; if(!silent && !htmlOutput){printf("Extending the overlapping alignments\n");}
}
if((strcmp(argv[i], "-forwardonly")) ==0){ //Consider forward alignments only
forwardAlignOnly = true;
if(!silent && !htmlOutput){printf("Considering forward direction alignments only\n");}
}
if((strcmp(argv[i], "-printpairwise")) ==0){
pairwiseOnly=true; if(!silent && !htmlOutput){printf("Printing pairwise scores only\n");}
}
if((strcmp(argv[i], "-FBP")) ==0){
FBP_on=true; if(!silent && !htmlOutput){printf("Using FBP profiles\n");}
}
if((strcmp(argv[i], "-useweighting")) ==0){
weighting_on=true; if(!silent && !htmlOutput){printf("Using weighting in FBP construction\n");}
}
if((strcmp(argv[i], "-prealigned")) ==0){
preAlign=true; if(!silent && !htmlOutput){printf("Profiles are pre-aligned\n");}
}
//Input TF dataset name
if((strcmp(argv[i], "-tf")) ==0)
{ if(argv[i+1]!=NULL)
{ strcpy(inputTFs, argv[i+1]);}
inputProvided=true;
}
//Score distribution file Make an auto function for this!!!!!!!
if((strcmp(argv[i], "-sd")) ==0)
{ if(argv[i+1]!=NULL)
{ strcpy(scoreDist, argv[i+1]);}
scoresProvided=true;
}
//Match input TFs against this dataset
if((strcmp(argv[i], "-match")) ==0)
{ if(argv[i+1]!=NULL)
{ strcpy(matchTFs, argv[i+1]);}
if(argv[i+2]!=NULL && strcmp(argv[i+2], "fams")==0){
famNames=true;
}
simMatching=true;
}
if((strcmp(argv[i], "-match_top")) ==0){ //Report the top X matches
if(argv[i+1]!=NULL)
{ matchTopX=strtol(argv[i+1], NULL, 10);}
}
//Matching input protein (Pfam) alignment dataset name
if((strcmp(argv[i], "-prot")) ==0)
{ if(argv[i+1]!=NULL)
{ strcpy(inputProteins, argv[i+1]);}
usingDomains = true;
}
//Run some tests
if((strcmp(argv[i], "-test")) ==0)
{ testing=true;
}
//Run some different tests
if((strcmp(argv[i], "-testacc")) ==0)
{ testingAcc=true;
famNames=true;
}
//Run some tests with trees
if((strcmp(argv[i], "-testtree")) ==0)
{ testingTree=true;
famNames=true;
}//Run Calinski & Harabasz with trees
if((strcmp(argv[i], "-ch")) ==0)
{ testingTree=true; treeClusts=true;
}//Run Calinski & Harabasz with trees and print the resulting clusters
if((strcmp(argv[i], "-chp")) ==0)
{ testingTree=true;
printTreeClusts=true; treeClusts=true;
}
}
//Defaults
if(!colChosen)
{ CC = new PearsonCorrelation();}
//Second Pass
for(i=1; i<argc; i++)
{
if((strcmp(argv[i], "-align")) ==0) //Choose an alignment method
{
if((strcmp(argv[i+1], "NW"))==0 || (strcmp(argv[i+1], "nw"))==0){
ALIGN = new NeedlemanWunsch(CC, gapOpen, gapExtend, overlapAlign, extendOverlap, forwardAlignOnly);
}
if((strcmp(argv[i+1], "SWU"))==0 || (strcmp(argv[i+1], "swu"))==0){
ALIGN = new SmithWatermanUngappedExtended(CC,forwardAlignOnly); ungapped=true;
}
if((strcmp(argv[i+1], "SWA"))==0 || (strcmp(argv[i+1], "swa"))==0){
ALIGN = new SmithWatermanAffine(CC, gapOpen, gapExtend, overlapAlign, extendOverlap,forwardAlignOnly);
}
if((strcmp(argv[i+1], "SW"))==0 || (strcmp(argv[i+1], "sw"))==0){
ALIGN = new SmithWaterman(CC, gapOpen, gapExtend, overlapAlign, extendOverlap,forwardAlignOnly);
}
alignChosen = true;
}
//Choose a multiple alignment method
if((strcmp(argv[i], "-ma")) ==0)
{
if((strcmp(argv[i+1], "PPA"))==0 || (strcmp(argv[i+1], "ppa"))==0){
MA = new ProgressiveProfileAlignment(outFileName, htmlOutput);
maChosen=true;
}
if((strcmp(argv[i+1], "IR"))==0 || (strcmp(argv[i+1], "ir"))==0){
MA = new IterativeRefinementAlignment(outFileName, htmlOutput);
maChosen=true;
}
if((strcmp(argv[i+1], "NONE"))==0 || (strcmp(argv[i+1], "none"))==0){
maChosen=true; ma_off=true;
}
}
}
if(!alignChosen)
{ ALIGN = new SmithWatermanAffine(CC, gapOpen, gapExtend, overlapAlign, extendOverlap);
}
if(!maChosen)
MA = new ProgressiveProfileAlignment(outFileName, htmlOutput);
//Third pass
//Choose a tree-construction method
for(i=1; i<argc; i++)
{ if((strcmp(argv[i], "-tree")) ==0)
{
if((strcmp(argv[i+1], "UPGMA"))==0 || (strcmp(argv[i+1], "upgma"))==0){
T = new UPGMA(ALIGN);
}
if((strcmp(argv[i+1], "SOTA"))==0 || (strcmp(argv[i+1], "sota"))==0){
T = new SOTA(ALIGN, MA); neuralTree=true;
}
if((strcmp(argv[i+1], "NJ"))==0 || (strcmp(argv[i+1], "nj"))==0){
T = new Neighbourjoin(ALIGN); printf("Using Neighbour-joining... ensure that the distance metric is additive\n");
}
if((strcmp(argv[i+1], "TDHC"))==0 || (strcmp(argv[i+1], "tdhc"))==0){
T = new TopDownHClust(ALIGN, MA); neuralTree=true;
}
treeChosen=true;
}
}
if(!treeChosen)
T = new UPGMA(ALIGN);
T->BeQuiet(silent);
////////////////////////////////////////////////////////////////////////////////////
//////// Main Program /////////////////////////////////////////////////////////////
//Initialise the background
Plat->ReadBackground();
if(inputProvided){
//Read in the matrices
Plat->ReadTransfacFile(inputTFs, famNames,true, weighting_on);
if(!silent && !htmlOutput){
printf("MatCount: %d\n", Plat->GetMatCount());
if(ungapped)
printf("Ungapped Alignment\n");
else
printf("Gap open = %.3lf, gap extend = %.3lf\n", gapOpen, gapExtend);
}
}else{
printf("No input motifs provided!\n\n");
}
if(genRandMotifs){
//Generate some random matrices
RandPSSMGen* RPG = new RandPSSMGen(Plat->inputMotifs, Plat->GetMatCount(), 10000, randMatOut);
RPG->RunGenerator();
}
if(genRandScores){
//Find rand dist
Plat->GetRandDistrib(scoresOut, ALIGN);
}else if(!scoresProvided){
printf("No score distribution file provided!\n\n");
}
if(testing){
PlatformTesting* PT = new PlatformTesting(CC);
//Print the distribution of column depth
// PT->ColumnDepthDist(Plat->inputMotifs, Plat->GetMatCount());
//Print the similarities of all columns against all others
// PT->ColumnScoreDist(Plat->inputMotifs, Plat->GetMatCount(), 0.05);
double z;
for(z=0.25; z<0.8; z+=0.05)
PT->RandColumns(Plat, z);
for(z=0.8; z<=1.0; z+=0.01)
PT->RandColumns(Plat, z);
delete(PT);
}
if(scoresProvided || preAlign){
Plat->ReadScoreDists(scoreDist);
if(!silent && !htmlOutput){printf("Scores read\n");}
if(Plat->GetMatCount()>1){
if(preAlign){
//No alignments or trees built here
pssmAlignment = MA->PreAlignedInput(Plat);
}else{
//Multiple alignment procedure
Plat->PreAlign(ALIGN);
if(pairwiseOnly){
if(!silent && !htmlOutput){printf("\nPairwise alignment scores:\n");}
Plat->PrintPairwise();
}if(!ma_off){
MA->ImportBasics(Plat, ALIGN);
if(!silent && !htmlOutput){printf("Alignments Finished\n");}
if(!testingAcc){
if(tree_loocv && testingTree){
T->LOOCVBuildTree(Plat, testingTree);
}else{
if(testingTree && !silent && !htmlOutput){printf("Calinski & Harabasz:\n\tNumClust\tC&H_Metric\n");}
T->BuildTree(Plat, testingTree);
if(!silent && treeClusts){printf("The Calinski & Harabasz statistic suggests %.0lf clusters in the input motifs\n", T->GetNodesMinCH());}
if(printTreeClusts){
T->PrintLevel(outFileName, int(T->GetNodesMinCH()));
}
}
T->PrintTree(outFileName);
if(!silent && !htmlOutput){printf("Tree Built\n");}
if(!silent){
if(!silent && !htmlOutput){printf("Multiple Alignment:\n");}
pssmAlignment = MA->BuildAlignment(Plat, ALIGN, T);
}
}
}
}
//Experiment with the Protein Domains
if(usingDomains){
PROTS = new ProteinDomains();
PROTS->ReadDomains(inputProteins, Plat->inputMotifs, Plat->GetMatCount());
PROTS->MutualInformation(pssmAlignment, MA->Alignment2Profile(pssmAlignment, "AlignmentMotif"), Plat->inputMotifs, Plat->GetMatCount());
delete PROTS;
}
}
//Similarity match against the database
if(simMatching){
Plat->ReadTransfacFile(matchTFs, famNames, false, false);
Plat->SimilarityMatching(ALIGN, outFileName, famNames, matchTopX);
}
}
if(testingAcc && scoresProvided && inputProvided && Plat->GetMatCount()>1){
PlatformTesting* PT = new PlatformTesting(CC);
PT->PairwisePredictionAccuracy(Plat);
}
delete(MA);
delete(T);
delete(CC);
delete(ALIGN);
}
delete(Plat);
return(0);
}
void generate_errors_per_base(JSONWriter* pWriter, const BWTIndexSet& index_set)
{
int n_samples = 100000;
size_t k = 25;
double max_error_rate = 0.95;
size_t min_overlap = 50;
std::vector<size_t> position_count;
std::vector<size_t> error_count;
Timer timer("test", true);
#if HAVE_OPENMP
omp_set_num_threads(opt::numThreads);
#pragma omp parallel for
#endif
for(int i = 0; i < n_samples; ++i)
{
std::string s = BWTAlgorithms::sampleRandomString(index_set.pBWT);
KmerOverlaps::retrieveMatches(s, k, min_overlap, max_error_rate, 2, index_set);
//KmerOverlaps::approximateMatch(s, min_overlap, max_error_rate, 2, 200, index_set);
MultipleAlignment ma =
KmerOverlaps::buildMultipleAlignment(s, k, min_overlap, max_error_rate, 2, index_set);
// Skip when there is insufficient depth to classify errors
size_t ma_rows = ma.getNumRows();
if(ma_rows <= 1)
continue;
size_t ma_cols = ma.getNumColumns();
size_t position = 0;
for(size_t j = 0; j < ma_cols; ++j)
{
char s_symbol = ma.getSymbol(0, j);
// Skip gaps
if(s_symbol == '-' || s_symbol == '\0')
continue;
SymbolCountVector scv = ma.getSymbolCountVector(j);
int s_symbol_count = 0;
char max_symbol = 0;
int max_count = 0;
for(size_t k = 0; k < scv.size(); ++k)
{
if(scv[k].symbol == s_symbol)
s_symbol_count = scv[k].count;
if(scv[k].count > max_count)
{
max_count = scv[k].count;
max_symbol = scv[k].symbol;
}
}
//printf("P: %zu S: %c M: %c MC: %d\n", position, s_symbol, max_symbol, max_count);
// Call an error at this position if the consensus symbol differs from the read
// and the support for the read symbol is less than 4 and the consensus symbol
// is strongly supported.
bool is_error = s_symbol != max_symbol && s_symbol_count < 4 && max_count >= 3;
#if HAVE_OPENMP
#pragma omp critical
#endif
{
if(position >= position_count.size())
{
position_count.resize(position+1);
error_count.resize(position+1);
}
position_count[position]++;
error_count[position] += is_error;
}
position += 1;
}
}
pWriter->String("ErrorsPerBase");
pWriter->StartObject();
pWriter->String("base_count");
pWriter->StartArray();
for(size_t i = 0; i < position_count.size(); ++i)
pWriter->Int(position_count[i]);
pWriter->EndArray();
pWriter->String("error_count");
pWriter->StartArray();
for(size_t i = 0; i < position_count.size(); ++i)
pWriter->Int(error_count[i]);
pWriter->EndArray();
pWriter->EndObject();
}
