void ClipOverlap::cleanupMateMap(MateMapByCoord& mateMap,
SamCoordOutput* outputBufferPtr,
int32_t chrom, int32_t position)
{
// Cleanup any reads in the mateMap whose mates are prior to the position
// currently being processed in the file. It means the mate was not found
// as expected. Stop cleaning up once one is found that is not passed.
uint64_t chromPos = 0;
if((chrom != -1) && (position != -1))
{
chromPos = SamHelper::combineChromPos(chrom, position);
}
else
{
chrom = -1;
}
// Stop after the first read is found whose mate has not yet been reached.
SamRecord* firstRec = mateMap.first();
while(firstRec != NULL)
{
uint64_t firstMateChromPos =
SamHelper::combineChromPos(firstRec->getMateReferenceID(),
firstRec->get0BasedMatePosition());
if((firstMateChromPos < chromPos) || (chrom == -1))
{
// Already past the mate's position, so note this read and
// write it.
++myNumMateFailures;
if((outputBufferPtr != NULL) && !myOverlapsOnly)
{
outputBufferPtr->add(firstRec);
}
else
{
myPool.releaseRecord(firstRec);
}
// Remove this record.
mateMap.popFirst();
// Get the next record to check.
firstRec = mateMap.first();
}
else
{
// The first record's mate position has not yet been passed, so
// stop cleaning up the buffer.
break;
}
}
}
// When a record is read, check if it is a duplicate or
// store for future checking.
void Dedup_LowMem::checkDups(SamRecord& record, uint32_t recordCount)
{
// Only inside this method if the record is mapped.
// Get the key for this record.
static DupKey key;
key.initKey(record, getLibraryID(record));
int flag = record.getFlag();
bool recordPaired = SamFlag::isPaired(flag) && SamFlag::isMateMapped(flag);
int sumBaseQual = getBaseQuality(record);
int32_t chromID = record.getReferenceID();
int32_t mateChromID = record.getMateReferenceID();
// If we are one-chrom and the mate is not on the same chromosome,
// mark it as not paired.
if(myOneChrom && (chromID != mateChromID))
{
recordPaired = false;
}
// Look in the fragment map to see if an entry for this key exists.
FragmentMapInsertReturn ireturn =
myFragmentMap.insert(std::make_pair(key, FragData()));
FragData* fragData = &(ireturn.first->second);
// Enter the new record in the fragData if (any of the below):
// 1) there is no previous entry for this key (ireturn.second == true)
// or
// 2) the previous entry is not paired
// AND
// a) the new record is paired
// or
// b) the new record has higher quality
if((ireturn.second == true) ||
((fragData->paired == false) &&
(recordPaired || (sumBaseQual > fragData->sumBaseQual))))
{
// Check if this is a new key.
if(ireturn.second == true)
{
// New entry, so build the recalibration table now.
if(myDoRecab)
{
myRecab.processReadBuildTable(record);
}
}
else if(fragData->paired == false)
{
// There was a previous record and it is not paired,
// so mark it as a duplicate.
// Duplicate checking/marking for pairs is handled below.
handleDuplicate(fragData->recordIndex);
}
// Store this record for later duplicate checking.
fragData->sumBaseQual = sumBaseQual;
fragData->recordIndex = recordCount;
fragData->paired = recordPaired;
}
else
{
// Leave the old record in fragData.
// If the new record is not paired, handle it as a duplicate.
if(recordPaired == false)
{
// This record is a duplicate, so mark it and release it.
handleDuplicate(recordCount);
}
}
// Only paired processing is left, so return if not paired.
if(recordPaired == false)
{
// Not paired, no more operations required, so return.
return;
}
// This is a paired record, so check for its mate.
uint64_t readPos =
SamHelper::combineChromPos(chromID,
record.get0BasedPosition());
uint64_t matePos =
SamHelper::combineChromPos(mateChromID,
record.get0BasedMatePosition());
int mateIndex = -1;
MateData* mateData = NULL;
// Check to see if the mate is prior to this record.
if(matePos <= readPos)
{
// The mate map is stored by the mate position, so look for this
// record's position.
// The mate should be in the mate map, so find it.
std::pair<MateMap::iterator,MateMap::iterator> matches =
myMateMap.equal_range(readPos);
// Loop through the elements that matched the pos looking for the mate.
for(MateMap::iterator iter = matches.first;
iter != matches.second; iter++)
{
if(strcmp((*iter).second.readName.c_str(),
record.getReadName()) == 0)
{
// Found the match.
mateData = &((*iter).second);
// Update the quality and track the mate record and index.
sumBaseQual += mateData->sumBaseQual;
mateIndex = mateData->recordIndex;
// Remove the entry from the map.
myMateMap.erase(iter);
break;
}
}
}
if(mateData == NULL)
{
if(matePos >= readPos)
{
// Haven't gotten to the mate yet, so store this record.
MateMap::iterator mateIter =
myMateMap.insert(std::make_pair(matePos, MateData()));
mateIter->second.sumBaseQual = sumBaseQual;
mateIter->second.recordIndex = recordCount;
mateIter->second.key.copy(key);
mateIter->second.readName = record.getReadName();
}
else
{
// Passed the mate, but it was not found.
handleMissingMate(record.getReferenceID(), record.getMateReferenceID());
}
return;
}
// Make the paired key.
PairedKey pkey(key, mateData->key);
// Check to see if this pair is a duplicate.
PairedMapInsertReturn pairedReturn =
myPairedMap.insert(std::make_pair(pkey,PairedData()));
PairedData* storedPair = &(pairedReturn.first->second);
// Get the index for "record 1" - the one with the earlier coordinate.
int record1Index = getFirstIndex(key, recordCount,
mateData->key, mateIndex);
// Check if we have already found a duplicate pair.
// If there is no duplicate found, there is nothing more to do.
if(pairedReturn.second == false)
{
// Duplicate found.
bool keepStored = true;
if(pairedReturn.first->second.sumBaseQual < sumBaseQual)
{
// The new pair has higher quality, so keep that.
keepStored = false;
}
else if(pairedReturn.first->second.sumBaseQual == sumBaseQual)
{
// Same quality, so keep the one with the earlier record1Index.
if(record1Index < storedPair->record1Index)
{
// The new pair has an earlier lower coordinate read,
// so keep that.
keepStored = false;
}
}
// Check to see which one should be kept by checking qualities.
if(keepStored)
{
// The old pair had higher quality so mark the new pair as a
// duplicate and release them.
handleDuplicate(mateIndex);
handleDuplicate(recordCount);
}
else
{
// The new pair has higher quality, so keep that.
// First mark the previous one as duplicates and release them.
handleDuplicate(storedPair->record1Index);
handleDuplicate(storedPair->record2Index);
// Store this pair's information.
if(record1Index == mateIndex)
{
// Mate has a lower coordinate, so make mate
// record1.
storedPair->sumBaseQual = sumBaseQual;
storedPair->record1Index = mateIndex;
storedPair->record2Index = recordCount;
}
else
{
// This record has a lower coordinate, so make it
// record1.
storedPair->sumBaseQual = sumBaseQual;
storedPair->record1Index = recordCount;
storedPair->record2Index = mateIndex;
}
}
}
else
{
// Store this pair's information.
storedPair->sumBaseQual = sumBaseQual;
if(record1Index == mateIndex)
{
// Mate has a lower coordinate, so make mate
// record1.
storedPair->record1Index = mateIndex;
storedPair->record2Index = recordCount;
}
else
{
// This record has a lower coordinate, so make it
// record1.
storedPair->record1Index = recordCount;
storedPair->record2Index = mateIndex;
}
}
}
void Bam2FastQ::handlePairedCoord(SamRecord& samRec)
{
static uint64_t readPos;
static uint64_t matePos;
static SamRecord* mateRec;
// This is a paired record, so check for its mate.
readPos = SamHelper::combineChromPos(samRec.getReferenceID(),
samRec.get0BasedPosition());
matePos = SamHelper::combineChromPos(samRec.getMateReferenceID(),
samRec.get0BasedMatePosition());
// Check to see if the mate is prior to this record.
if(matePos <= readPos)
{
// The mate is prior to this position, so look for this record.
mateRec = myMateMap.getMate(samRec);
if(mateRec == NULL)
{
// If they are the same position, add it to the map.
if(matePos == readPos)
{
myMateMap.add(samRec);
// Check to see if the mate map can be cleaned up prior
// to this position.
cleanUpMateMap(readPos);
}
else
{
// Paired Read, but could not find mate.
std::cerr << "Paired Read, " << samRec.getReadName()
<< " but couldn't find mate, so writing as "
<< "unpaired (single-ended)\n";
++myNumMateFailures;
writeFastQ(samRec, myUnpairedFile, myUnpairedFileNameExt);
}
}
else
{
// Found the mate.
++myNumPairs;
// Check which is the first in the pair.
if(SamFlag::isFirstFragment(samRec.getFlag()))
{
if(SamFlag::isFirstFragment(mateRec->getFlag()))
{
std::cerr << "Both reads of " << samRec.getReadName()
<< " are first fragment, so "
<< "splitting one to be in the 2nd fastq.\n";
}
writeFastQ(samRec, myFirstFile, myFirstFileNameExt, myFirstRNExt.c_str());
writeFastQ(*mateRec, mySecondFile, mySecondFileNameExt, mySecondRNExt.c_str());
}
else
{
if(!SamFlag::isFirstFragment(mateRec->getFlag()))
{
std::cerr << "Neither read of " << samRec.getReadName()
<< " are first fragment, so "
<< "splitting one to be in the 2nd fastq.\n";
}
writeFastQ(*mateRec, myFirstFile, myFirstFileNameExt, myFirstRNExt.c_str());
writeFastQ(samRec, mySecondFile, mySecondFileNameExt, mySecondRNExt.c_str());
}
}
}
else
{
// Haven't gotten to the mate yet, so store it.
myMateMap.add(samRec);
// Check to see if the mate map can be cleaned up.
cleanUpMateMap(readPos);
}
}
void validateRead1ModQuality(SamRecord& samRecord)
{
//////////////////////////////////////////
// Validate Record 1
// Create record structure for validating.
int expectedBlockSize = 89;
const char* expectedReferenceName = "1";
const char* expectedMateReferenceName = "1";
const char* expectedMateReferenceNameOrEqual = "=";
bamRecordStruct* expectedRecordPtr =
(bamRecordStruct *) malloc(expectedBlockSize + sizeof(int));
char tag[3];
char type;
void* value;
bamRecordStruct* bufferPtr;
unsigned char* varPtr;
expectedRecordPtr->myBlockSize = expectedBlockSize;
expectedRecordPtr->myReferenceID = 0;
expectedRecordPtr->myPosition = 1010;
expectedRecordPtr->myReadNameLength = 23;
expectedRecordPtr->myMapQuality = 0;
expectedRecordPtr->myBin = 4681;
expectedRecordPtr->myCigarLength = 2;
expectedRecordPtr->myFlag = 73;
expectedRecordPtr->myReadLength = 5;
expectedRecordPtr->myMateReferenceID = 0;
expectedRecordPtr->myMatePosition = 1010;
expectedRecordPtr->myInsertSize = 0;
// Check the alignment end
assert(samRecord.get0BasedAlignmentEnd() == 1016);
assert(samRecord.get1BasedAlignmentEnd() == 1017);
assert(samRecord.getAlignmentLength() == 7);
assert(samRecord.get0BasedUnclippedStart() == 1010);
assert(samRecord.get1BasedUnclippedStart() == 1011);
assert(samRecord.get0BasedUnclippedEnd() == 1016);
assert(samRecord.get1BasedUnclippedEnd() == 1017);
// Check the accessors.
assert(samRecord.getBlockSize() == expectedRecordPtr->myBlockSize);
assert(samRecord.getReferenceID() == expectedRecordPtr->myReferenceID);
assert(strcmp(samRecord.getReferenceName(), expectedReferenceName) == 0);
assert(samRecord.get1BasedPosition() == expectedRecordPtr->myPosition + 1);
assert(samRecord.get0BasedPosition() == expectedRecordPtr->myPosition);
assert(samRecord.getReadNameLength() ==
expectedRecordPtr->myReadNameLength);
assert(samRecord.getMapQuality() == expectedRecordPtr->myMapQuality);
assert(samRecord.getBin() == expectedRecordPtr->myBin);
assert(samRecord.getCigarLength() == expectedRecordPtr->myCigarLength);
assert(samRecord.getFlag() == expectedRecordPtr->myFlag);
assert(samRecord.getReadLength() == expectedRecordPtr->myReadLength);
assert(samRecord.getMateReferenceID() ==
expectedRecordPtr->myMateReferenceID);
assert(strcmp(samRecord.getMateReferenceName(),
expectedMateReferenceName) == 0);
assert(strcmp(samRecord.getMateReferenceNameOrEqual(),
expectedMateReferenceNameOrEqual) == 0);
assert(samRecord.get1BasedMatePosition() ==
expectedRecordPtr->myMatePosition + 1);
assert(samRecord.get0BasedMatePosition() ==
expectedRecordPtr->myMatePosition);
assert(samRecord.getInsertSize() == expectedRecordPtr->myInsertSize);
assert(strcmp(samRecord.getReadName(), "1:1011:F:255+17M15D20M") == 0);
assert(strcmp(samRecord.getCigar(), "5M2D") == 0);
assert(strcmp(samRecord.getSequence(), "CCGAA") == 0);
assert(strcmp(samRecord.getQuality(), "ABCDE") == 0);
assert(samRecord.getNumOverlaps(1010, 1017) == 5);
assert(samRecord.getNumOverlaps(1010, 1016) == 5);
assert(samRecord.getNumOverlaps(1012, 1017) == 3);
assert(samRecord.getNumOverlaps(1015, 1017) == 0);
assert(samRecord.getNumOverlaps(1017, 1010) == 0);
assert(samRecord.getNumOverlaps(1013, 1011) == 0);
assert(samRecord.getNumOverlaps(-1, 1017) == 5);
// Reset the tag iter, since the tags have already been read.
samRecord.resetTagIter();
// Check the tags.
assert(samRecord.getNextSamTag(tag, type, &value) == true);
assert(tag[0] == 'A');
assert(tag[1] == 'M');
assert(type == 'i');
assert(*(char*)value == 0);
assert(samRecord.getNextSamTag(tag, type, &value) == true);
assert(tag[0] == 'M');
assert(tag[1] == 'D');
assert(type == 'Z');
assert(*(String*)value == "37");
assert(samRecord.getNextSamTag(tag, type, &value) == true);
assert(tag[0] == 'N');
assert(tag[1] == 'M');
assert(type == 'i');
assert(*(char*)value == 0);
assert(samRecord.getNextSamTag(tag, type, &value) == true);
assert(tag[0] == 'X');
assert(tag[1] == 'T');
assert(type == 'A');
assert(*(char*)value == 'R');
// No more tags, should return false.
assert(samRecord.getNextSamTag(tag, type, &value) == false);
assert(samRecord.getNextSamTag(tag, type, &value) == false);
// Get the record ptr.
bufferPtr = (bamRecordStruct*)samRecord.getRecordBuffer();
// Validate the buffers match.
assert(bufferPtr->myBlockSize == expectedRecordPtr->myBlockSize);
assert(bufferPtr->myReferenceID == expectedRecordPtr->myReferenceID);
assert(bufferPtr->myPosition == expectedRecordPtr->myPosition);
assert(bufferPtr->myReadNameLength == expectedRecordPtr->myReadNameLength);
assert(bufferPtr->myMapQuality == expectedRecordPtr->myMapQuality);
assert(bufferPtr->myBin == expectedRecordPtr->myBin);
assert(bufferPtr->myCigarLength == expectedRecordPtr->myCigarLength);
assert(bufferPtr->myFlag == expectedRecordPtr->myFlag);
assert(bufferPtr->myReadLength == expectedRecordPtr->myReadLength);
assert(bufferPtr->myMateReferenceID ==
expectedRecordPtr->myMateReferenceID);
assert(bufferPtr->myMatePosition == expectedRecordPtr->myMatePosition);
assert(bufferPtr->myInsertSize == expectedRecordPtr->myInsertSize);
// Validate the variable length fields in the buffer.
// Set the pointer to the start of the variable fields.
varPtr = (unsigned char*)(&(bufferPtr->myData[0]));
// Validate the readname.
for(int i = 0; i < expectedRecordPtr->myReadNameLength; i++)
{
assert(*varPtr == samRecord.getReadName()[i]);
varPtr++;
}
// Validate the cigar.
// The First cigar is 5M which is 5 << 4 | 0 = 80
assert(*(unsigned int*)varPtr == 80);
// Increment the varptr the size of an int.
varPtr += 4;
// The 2nd cigar is 2D which is 2 << 4 | 2 = 34
assert(*(unsigned int*)varPtr == 34);
// Increment the varptr the size of an int.
varPtr += 4;
// Validate the sequence.
// CC = 0x22
assert(*varPtr == 0x22);
varPtr++;
// GA = 0x41
assert(*varPtr == 0x41);
varPtr++;
// A = 0x10
assert(*varPtr == 0x10);
varPtr++;
// Validate the Quality
for(int i = 0; i < expectedRecordPtr->myReadLength; i++)
{
assert(*varPtr == samRecord.getQuality()[i] - 33);
varPtr++;
}
// Validate the tags.
assert(*varPtr == 'A');
varPtr++;
assert(*varPtr == 'M');
varPtr++;
assert(*varPtr == 'C');
varPtr++;
assert(*varPtr == 0);
varPtr++;
assert(*varPtr == 'M');
varPtr++;
assert(*varPtr == 'D');
varPtr++;
assert(*varPtr == 'Z');
varPtr++;
assert(*varPtr == '3');
varPtr++;
assert(*varPtr == '7');
varPtr++;
assert(*varPtr == 0);
varPtr++;
assert(*varPtr == 'N');
varPtr++;
assert(*varPtr == 'M');
varPtr++;
assert(*varPtr == 'C');
varPtr++;
assert(*varPtr == 0);
varPtr++;
assert(*varPtr == 'X');
varPtr++;
assert(*varPtr == 'T');
varPtr++;
assert(*varPtr == 'A');
varPtr++;
assert(*varPtr == 'R');
varPtr++;
}
int GapInfo::processFile(const char* inputFileName, const char* outputFileName,
const char* refFile, bool detailed,
bool checkFirst, bool checkStrand)
{
// Open the file for reading.
SamFile samIn;
samIn.OpenForRead(inputFileName);
// Read the sam header.
SamFileHeader samHeader;
samIn.ReadHeader(samHeader);
SamRecord samRecord;
GenomeSequence* refPtr = NULL;
if(strcmp(refFile, "") != 0)
{
refPtr = new GenomeSequence(refFile);
}
IFILE outFile = ifopen(outputFileName, "w");
// Map for summary.
std::map<int, int> gapInfoMap;
// Keep reading records until ReadRecord returns false.
while(samIn.ReadRecord(samHeader, samRecord))
{
uint16_t samFlags = samRecord.getFlag();
if((!SamFlag::isMapped(samFlags)) ||
(!SamFlag::isMateMapped(samFlags)) ||
(!SamFlag::isPaired(samFlags)) ||
(samFlags & SamFlag::SECONDARY_ALIGNMENT) ||
(SamFlag::isDuplicate(samFlags)) ||
(SamFlag::isQCFailure(samFlags)))
{
// unmapped, mate unmapped, not paired,
// not the primary alignment,
// duplicate, fails vendor quality check
continue;
}
// No gap info if the chromosome names are different or
// are unknown.
int32_t refID = samRecord.getReferenceID();
if((refID != samRecord.getMateReferenceID()) || (refID == -1))
{
continue;
}
int32_t readStart = samRecord.get0BasedPosition();
int32_t mateStart = samRecord.get0BasedMatePosition();
// If the mate starts first, then the pair was processed by
// the mate.
if(mateStart < readStart)
{
continue;
}
if((mateStart == readStart) && (SamFlag::isReverse(samFlags)))
{
// read and mate start at the same position, so
// only process the forward strand.
continue;
}
// Process this read pair.
int32_t readEnd = samRecord.get0BasedAlignmentEnd();
int32_t gapSize = mateStart - readEnd - 1;
if(detailed)
{
// Output the gap info.
ifprintf(outFile, "%s\t%d\t%d",
samRecord.getReferenceName(), readEnd+1, gapSize);
// Check if it is not the first or if it is not the forward strand.
if(checkFirst && !SamFlag::isFirstFragment(samFlags))
{
ifprintf(outFile, "\tNotFirst");
}
if(checkStrand && SamFlag::isReverse(samFlags))
{
ifprintf(outFile, "\tReverse");
}
ifprintf(outFile, "\n");
}
else
{
// Summary.
// Skip reads that are not the forward strand.
if(SamFlag::isReverse(samFlags))
{
// continue
continue;
}
// Forward.
// Check the reference for 'N's.
if(refPtr != NULL)
{
genomeIndex_t chromStartIndex =
refPtr->getGenomePosition(samRecord.getReferenceName());
if(chromStartIndex == INVALID_GENOME_INDEX)
{
// Invalid position, so continue to the next one.
continue;
}
bool skipRead = false;
for(int i = readEnd + 1; i < mateStart; i++)
{
if((*refPtr)[i] == 'N')
{
// 'N' in the reference, so continue to the next read.
skipRead = true;
break;
}
}
if(skipRead)
{
continue;
}
}
// Update the gapInfo.
gapInfoMap[gapSize]++;
}
}
if(!detailed)
{
// Output the summary.
ifprintf(outFile, "GapSize\tNumPairs\n");
for(std::map<int,int>::iterator iter = gapInfoMap.begin();
iter != gapInfoMap.end(); iter++)
{
ifprintf(outFile, "%d\t%d\n", (*iter).first, (*iter).second);
}
}
SamStatus::Status returnStatus = samIn.GetStatus();
if(returnStatus == SamStatus::NO_MORE_RECS)
{
return(SamStatus::SUCCESS);
}
return(returnStatus);
}
// When a record is read, check if it is a duplicate or
// store for future checking.
void Dedup::checkDups(SamRecord& record, uint32_t recordCount)
{
// Only inside this method if the record is mapped.
// Get the key for this record.
static DupKey key;
static DupKey mateKey;
key.updateKey(record, getLibraryID(record));
int flag = record.getFlag();
bool recordPaired = SamFlag::isPaired(flag) && SamFlag::isMateMapped(flag);
int sumBaseQual = getBaseQuality(record);
int32_t chromID = record.getReferenceID();
int32_t mateChromID = record.getMateReferenceID();
// If we are one-chrom and the mate is not on the same chromosome,
// mark it as not paired.
if(myOneChrom && (chromID != mateChromID))
{
recordPaired = false;
}
// Look in the map to see if an entry for this key exists.
FragmentMapInsertReturn ireturn =
myFragmentMap.insert(std::make_pair(key, ReadData()));
ReadData* readData = &(ireturn.first->second);
// Mark this record's data in the fragment record if this is the first
// entry or if it is a duplicate and the old record is not paired and
// the new record is paired or the has a higher quality.
if((ireturn.second == true) ||
((readData->paired == false) &&
(recordPaired || (sumBaseQual > readData->sumBaseQual))))
{
// If there was a previous record, mark it duplicate and release
// the old record
if(ireturn.second == false)
{
// Mark the old record as a DUPLICATE!
handleDuplicate(readData->recordIndex, readData->recordPtr);
}
// Store this record for later duplicate checking.
readData->sumBaseQual = sumBaseQual;
readData->recordIndex = recordCount;
readData->paired = recordPaired;
if(recordPaired)
{
readData->recordPtr = NULL;
}
else
{
readData->recordPtr = &record;
}
}
else
{
// The old record is not a duplicate so the new record is
// a duplicate if it is not paired.
if(recordPaired == false)
{
// This record is a duplicate, so mark it and release it.
handleDuplicate(recordCount, &record);
}
}
// Only paired processing is left, so return if not paired.
if(recordPaired == false)
{
// Not paired, no more operations required, so return.
return;
}
// This is a paired record, so check for its mate.
uint64_t readPos =
SamHelper::combineChromPos(chromID,
record.get0BasedPosition());
uint64_t matePos =
SamHelper::combineChromPos(mateChromID,
record.get0BasedMatePosition());
SamRecord* mateRecord = NULL;
int mateIndex = 0;
// Check to see if the mate is prior to this record.
if(matePos <= readPos)
{
// The mate map is stored by the mate position, so look for this
// record's position.
// The mate should be in the mate map, so find it.
std::pair<MateMap::iterator,MateMap::iterator> matches =
myMateMap.equal_range(readPos);
// Loop through the elements that matched the pos looking for the mate.
for(MateMap::iterator iter = matches.first;
iter != matches.second; iter++)
{
if(strcmp((*iter).second.recordPtr->getReadName(),
record.getReadName()) == 0)
{
// Found the match.
ReadData* mateData = &((*iter).second);
// Update the quality and track the mate record and index.
sumBaseQual += mateData->sumBaseQual;
mateIndex = mateData->recordIndex;
mateRecord = mateData->recordPtr;
// Remove the entry from the map.
myMateMap.erase(iter);
break;
}
}
}
if((mateRecord == NULL) && (matePos >= readPos))
{
// Haven't gotten to the mate yet, so store this record.
MateMap::iterator mateIter =
myMateMap.insert(std::make_pair(matePos, ReadData()));
mateIter->second.sumBaseQual = sumBaseQual;
mateIter->second.recordPtr = &record;
mateIter->second.recordIndex = recordCount;
// No more processing for this record is necessary.
return;
}
if(mateRecord == NULL)
{
// Passed the mate, but it was not found.
handleMissingMate(&record);
return;
}
// Make the paired key.
mateKey.updateKey(*mateRecord, getLibraryID(*mateRecord));
PairedKey pkey(key, mateKey);
// Check to see if this pair is a duplicate.
PairedMapInsertReturn pairedReturn =
myPairedMap.insert(std::make_pair(pkey,PairedData()));
PairedData* storedPair = &(pairedReturn.first->second);
// Get the index for "record 1" - the one with the earlier coordinate.
int record1Index = getFirstIndex(key, recordCount,
mateKey, mateIndex);
// Check if we have already found a duplicate pair.
// If there is no duplicate found, there is nothing more to do.
if(pairedReturn.second == false)
{
// Duplicate found.
bool keepStored = true;
if(pairedReturn.first->second.sumBaseQual < sumBaseQual)
{
// The new pair has higher quality, so keep that.
keepStored = false;
}
else if(pairedReturn.first->second.sumBaseQual == sumBaseQual)
{
// Same quality, so keep the one with the earlier record1Index.
if(record1Index < storedPair->record1Index)
{
// The new pair has an earlier lower coordinate read,
// so keep that.
keepStored = false;
}
}
// Check to see which one should be kept by checking qualities.
if(keepStored)
{
// The old pair had higher quality so mark the new pair as a
// duplicate and release them.
handleDuplicate(mateIndex, mateRecord);
handleDuplicate(recordCount, &record);
}
else
{
// The new pair has higher quality, so keep that.
// First mark the previous one as duplicates and release them.
handleDuplicate(storedPair->record1Index, storedPair->record1Ptr);
handleDuplicate(storedPair->record2Index, storedPair->record2Ptr);
// Store this pair's information.
if(record1Index == mateIndex)
{
// Mate has a lower coordinate, so make mate
// record1.
storedPair->sumBaseQual = sumBaseQual;
storedPair->record1Ptr = mateRecord;
storedPair->record2Ptr = &record;
storedPair->record1Index = mateIndex;
storedPair->record2Index = recordCount;
}
else
{
// This record has a lower coordinate, so make it
// record1.
storedPair->sumBaseQual = sumBaseQual;
storedPair->record1Ptr = &record;
storedPair->record2Ptr = mateRecord;
storedPair->record1Index = recordCount;
storedPair->record2Index = mateIndex;
}
}
}
else
{
// Store this pair's information.
storedPair->sumBaseQual = sumBaseQual;
if(record1Index == mateIndex)
{
// Mate has a lower coordinate, so make mate
// record1.
storedPair->record1Ptr = mateRecord;
storedPair->record2Ptr = &record;
storedPair->record1Index = mateIndex;
storedPair->record2Index = recordCount;
}
else
{
// This record has a lower coordinate, so make it
// record1.
storedPair->record1Ptr = &record;
storedPair->record2Ptr = mateRecord;
storedPair->record1Index = recordCount;
storedPair->record2Index = mateIndex;
}
}
}
