//
// Main
//
int overlapMain(int argc, char** argv)
{
parseOverlapOptions(argc, argv);
// Prepare the output ASQG file
assert(opt::outputType == OT_ASQG);
// Open output file
std::ostream* pASQGWriter = createWriter(opt::outFile);
// Build and write the ASQG header
ASQG::HeaderRecord headerRecord;
headerRecord.setOverlapTag(opt::minOverlap);
headerRecord.setErrorRateTag(opt::errorRate);
headerRecord.setInputFileTag(opt::readsFile);
headerRecord.setContainmentTag(true); // containments are always present
headerRecord.setTransitiveTag(!opt::bIrreducibleOnly);
headerRecord.write(*pASQGWriter);
// Compute the overlap hits
StringVector hitsFilenames;
// Determine which index files to use. If a target file was provided,
// use the index of the target reads
std::string indexPrefix;
if(!opt::prefix.empty())
indexPrefix = opt::prefix;
else
{
if(!opt::targetFile.empty())
indexPrefix = stripFilename(opt::targetFile);
else
indexPrefix = stripFilename(opt::readsFile);
}
BWT* pBWT = new BWT(indexPrefix + BWT_EXT, opt::sampleRate);
BWT* pRBWT = new BWT(indexPrefix + RBWT_EXT, opt::sampleRate);
OverlapAlgorithm* pOverlapper = new OverlapAlgorithm(pBWT, pRBWT,
opt::errorRate, opt::seedLength,
opt::seedStride, opt::bIrreducibleOnly);
pOverlapper->setExactModeOverlap(opt::errorRate <= 0.0001);
pOverlapper->setExactModeIrreducible(opt::errorRate <= 0.0001);
Timer* pTimer = new Timer(PROGRAM_IDENT);
pBWT->printInfo();
// Make a prefix for the temporary hits files
std::string outPrefix;
outPrefix = stripFilename(opt::readsFile);
if(!opt::targetFile.empty())
{
outPrefix.append(1, '.');
outPrefix.append(stripFilename(opt::targetFile));
}
if(opt::numThreads <= 1)
{
printf("[%s] starting serial-mode overlap computation\n", PROGRAM_IDENT);
computeHitsSerial(outPrefix, opt::readsFile, pOverlapper, opt::minOverlap, hitsFilenames, pASQGWriter);
}
else
{
printf("[%s] starting parallel-mode overlap computation with %d threads\n", PROGRAM_IDENT, opt::numThreads);
computeHitsParallel(opt::numThreads, outPrefix, opt::readsFile, pOverlapper, opt::minOverlap, hitsFilenames, pASQGWriter);
}
// Get the number of strings in the BWT, this is used to pre-allocated the read table
delete pOverlapper;
delete pBWT;
delete pRBWT;
// Parse the hits files and write the overlaps to the ASQG file
convertHitsToASQG(indexPrefix, hitsFilenames, pASQGWriter);
// Cleanup
delete pASQGWriter;
delete pTimer;
if(opt::numThreads > 1)
pthread_exit(NULL);
return 0;
}
void cluster()
{
BWT* pBWT = new BWT(opt::prefix + BWT_EXT);
BWT* pRBWT = new BWT(opt::prefix + RBWT_EXT);
OverlapAlgorithm* pOverlapper = new OverlapAlgorithm(pBWT, pRBWT,opt::errorRate, opt::seedLength, opt::seedStride, true);
pOverlapper->setExactModeOverlap(opt::errorRate < 0.001f);
pOverlapper->setExactModeIrreducible(opt::errorRate < 0.001f);
BitVector markedReads(pBWT->getNumStrings());
std::string preclustersFile = opt::outFile + ".preclusters";
std::ostream* pPreWriter = createWriter(preclustersFile);
ClusterPostProcess postProcessor(pPreWriter, opt::minSize, &markedReads);
// Set the cluster parameters
ClusterParameters parameters;
parameters.pOverlapper = pOverlapper;
parameters.minOverlap = opt::minOverlap;
parameters.maxClusterSize = opt::maxSize;
parameters.maxIterations = opt::maxIterations;
parameters.pMarkedReads = &markedReads;
// Read the limit kmer sequences, if provided
std::set<std::string>* pLimitKmers = NULL;
if(!opt::limitFile.empty())
{
// Read in the limit sequences
pLimitKmers = new std::set<std::string>;
readLimitKmers(pLimitKmers);
parameters.pLimitKmers = pLimitKmers;
parameters.limitK = opt::limitKmer;
}
else
{
parameters.pLimitKmers = NULL;
parameters.limitK = 0;
}
// Make pre-clusters from the reads
if(opt::numThreads <= 1)
{
printf("[%s] starting serial-mode read clustering\n", PROGRAM_IDENT);
ClusterProcess processor(parameters);
// If the extend file is empty, build new clusters
if(opt::extendFile.empty())
{
PROCESS_CLUSTER_SERIAL(opt::readsFile, &processor, &postProcessor);
}
else
{
// Process a set of preexisting clusters
ClusterReader clusterReader(opt::extendFile);
PROCESS_EXTEND_SERIAL(clusterReader, &processor, &postProcessor);
}
}
else
{
printf("[%s] starting parallel-mode read clustering computation with %d threads\n", PROGRAM_IDENT, opt::numThreads);
std::vector<ClusterProcess*> processorVector;
for(int i = 0; i < opt::numThreads; ++i)
{
ClusterProcess* pProcessor = new ClusterProcess(parameters);
processorVector.push_back(pProcessor);
}
if(opt::extendFile.empty())
{
PROCESS_CLUSTER_PARALLEL(opt::readsFile, processorVector, &postProcessor);
}
else
{
ClusterReader clusterReader(opt::extendFile);
PROCESS_EXTEND_PARALLEL(clusterReader, processorVector, &postProcessor);
}
for(size_t i = 0; i < processorVector.size(); ++i)
{
delete processorVector[i];
processorVector[i] = NULL;
}
}
delete pPreWriter;
delete pBWT;
delete pRBWT;
delete pOverlapper;
// Deallocate limit kmers
if(pLimitKmers != NULL)
delete pLimitKmers;
// Open the preclusters file and convert them to read names
SuffixArray* pFwdSAI = new SuffixArray(opt::prefix + SAI_EXT);
ReadInfoTable* pRIT = new ReadInfoTable(opt::readsFile, pFwdSAI->getNumStrings());
size_t seedIdx = 0;
std::istream* pPreReader = createReader(preclustersFile);
std::ostream* pClusterWriter = createWriter(opt::outFile);
std::string line;
while(getline(*pPreReader,line))
{
std::stringstream parser(line);
std::string clusterName;
std::string readSequence;
size_t clusterSize;
int64_t lowIdx;
int64_t highIdx;
parser >> clusterName >> clusterSize >> readSequence >> lowIdx >> highIdx;
if(lowIdx > highIdx)
{
// This is an extra read that is not present in the FM-index
// Output a record with a fake read ID
*pClusterWriter << clusterName << "\t" << clusterSize << "\tseed-" << seedIdx++ << "\t" << readSequence << "\n";
}
else
{
for(int64_t i = lowIdx; i <= highIdx; ++i)
{
const ReadInfo& targetInfo = pRIT->getReadInfo(pFwdSAI->get(i).getID());
std::string readName = targetInfo.id;
*pClusterWriter << clusterName << "\t" << clusterSize << "\t" << readName << "\t" << readSequence << "\n";
}
}
}
unlink(preclustersFile.c_str());
delete pFwdSAI;
delete pRIT;
delete pPreReader;
delete pClusterWriter;
}
//
// Main
//
int FMMergeMain(int argc, char** argv)
{
parseFMMergeOptions(argc, argv);
BWT* pBWT = new BWT(opt::prefix + BWT_EXT);
BWT* pRBWT = new BWT(opt::prefix + RBWT_EXT);
OverlapAlgorithm* pOverlapper = new OverlapAlgorithm(pBWT, pRBWT,0.0f, 0,0,true);
pOverlapper->setExactModeOverlap(true);
pOverlapper->setExactModeIrreducible(true);
Timer* pTimer = new Timer(PROGRAM_IDENT);
pBWT->printInfo();
// Construct a bitvector indicating what reads have been used
// All the processes read from this vector and only the post processor
// writes to it.
BitVector markedReads(pBWT->getNumStrings());
std::ostream* pWriter = createWriter(opt::outFile);
FMMergePostProcess postProcessor(pWriter, &markedReads);
if(opt::numThreads <= 1)
{
printf("[%s] starting serial-mode read merging\n", PROGRAM_IDENT);
FMMergeProcess processor(pOverlapper, opt::minOverlap, &markedReads);
SequenceProcessFramework::processSequencesSerial<SequenceWorkItem,
FMMergeResult,
FMMergeProcess,
FMMergePostProcess>(opt::readsFile, &processor, &postProcessor);
}
else
{
printf("[%s] starting parallel-mode read merging computation with %d threads\n", PROGRAM_IDENT, opt::numThreads);
std::vector<FMMergeProcess*> processorVector;
for(int i = 0; i < opt::numThreads; ++i)
{
FMMergeProcess* pProcessor = new FMMergeProcess(pOverlapper, opt::minOverlap, &markedReads);
processorVector.push_back(pProcessor);
}
SequenceProcessFramework::processSequencesParallel<SequenceWorkItem,
FMMergeResult,
FMMergeProcess,
FMMergePostProcess>(opt::readsFile, processorVector, &postProcessor);
for(size_t i = 0; i < processorVector.size(); ++i)
{
delete processorVector[i];
processorVector[i] = NULL;
}
}
// Check that every bit was set in the bit vector
size_t numSet = 0;
size_t numTotal = pBWT->getNumStrings();
for(size_t i = 0; i < numTotal; ++i)
{
if(markedReads.test(i))
++numSet;
}
// Get the number of strings in the BWT, this is used to pre-allocated the read table
delete pOverlapper;
delete pBWT;
delete pRBWT;
delete pWriter;
// Cleanup
delete pTimer;
if(opt::numThreads > 1)
pthread_exit(NULL);
return 0;
}