public String assembleContigs(
List<String> inputFiles,
String output,
String tempDir,
List<Feature> regions,
String prefix,
boolean checkForDupes,
ReAligner realigner,
CompareToReference2 c2r) {
if ((kmers.length == 0) || (kmers[0] < KmerSizeEvaluator.MIN_KMER)) {
KmerSizeEvaluator kmerEval = new KmerSizeEvaluator();
int kmer = kmerEval.identifyMinKmer(readLength, c2r, regions);
this.kmers = realigner.toKmerArray(kmer, readLength);
}
String contigs = "";
long start = System.currentTimeMillis();
int readCount = 0;
int minReadCount = Integer.MAX_VALUE;
// if c2r is null, this is the unaligned region.
boolean isAssemblyCandidate = c2r == null ? true : false;
try {
List<List<SAMRecord>> readsList = getReads(inputFiles, regions, realigner);
for (List<SAMRecord> reads : readsList) {
int candidateReadCount = 0;
for (SAMRecord read : reads) {
if (!isAssemblyCandidate && isAssemblyTriggerCandidate(read, c2r)) {
candidateReadCount++;
}
if (shouldSearchForSv && isSvCandidate(read)) {
svCandidates.add(
new Position(read.getMateReferenceName(), read.getMateAlignmentStart()));
}
}
if (candidateReadCount > minCandidateCount(reads.size(), regions.get(0))) {
isAssemblyCandidate = true;
}
if (reads.size() < minReadCount) {
minReadCount = reads.size();
}
readCount += reads.size();
}
StringBuffer readBuffer = new StringBuffer();
if (isAssemblyCandidate) {
int downsampleTarget = desiredNumberOfReads(regions);
char sampleId = 1;
for (List<SAMRecord> reads : readsList) {
// Default to always keep
double keepProbability = 1.1;
if (reads.size() > downsampleTarget) {
keepProbability = (double) downsampleTarget / (double) reads.size();
}
Random random = new Random(1);
for (SAMRecord read : reads) {
if (random.nextDouble() < keepProbability) {
readBuffer.append(sampleId);
readBuffer.append(read.getReadNegativeStrandFlag() ? "1" : "0");
if (read.getReadString().length() == readLength) {
readBuffer.append(read.getReadString());
readBuffer.append(read.getBaseQualityString());
} else {
StringBuffer basePadding = new StringBuffer();
StringBuffer qualPadding = new StringBuffer();
for (int i = 0; i < readLength - read.getReadString().length(); i++) {
basePadding.append('N');
qualPadding.append('!');
}
readBuffer.append(read.getReadString() + basePadding.toString());
readBuffer.append(read.getBaseQualityString() + qualPadding.toString());
}
}
}
// Make this set of reads eligible for GC
reads.clear();
sampleId += 1;
}
}
readsList.clear();
if (isAssemblyCandidate) {
for (int kmer : kmers) {
String outputFile = output + "_k" + kmer;
contigs =
assemble(
readBuffer.toString(),
outputFile,
prefix,
truncateOnRepeat ? 1 : 0,
maxContigs,
maxPathsFromRoot,
readLength,
kmer,
minKmerFrequency,
minBaseQuality,
minEdgeRatio,
isDebug ? 1 : 0,
maxNodes);
if (!contigs.equals("<REPEAT>")) {
break;
} else {
if (kmer >= readLength / 2 || kmer >= CYCLE_KMER_LENGTH_THRESHOLD) {
isCycleExceedingThresholdDetected = true;
}
}
}
} else {
// System.out.println("Skipping assembly for: " + prefix);
}
} catch (Exception e) {
e.printStackTrace();
throw new RuntimeException(e);
}
if (this.shouldSearchForSv) {
Collections.sort(this.svCandidates);
Position last = null;
String currentFeatureChr = null;
int currentFeatureStart = -1;
int currentFeatureStop = -1;
int currentFeatureCount = 0;
// TODO: Calc this dynamically
int windowSize = 500;
for (Position pos : this.svCandidates) {
if ((last != null)
&& pos.getChromosome().equals(last.getChromosome())
&& Math.abs(pos.getPosition() - last.getPosition()) < windowSize) {
if (currentFeatureChr == null) {
currentFeatureChr = pos.getChromosome();
currentFeatureStart = last.getPosition();
currentFeatureStop = pos.getPosition() + readLength;
currentFeatureCount = 1;
} else {
currentFeatureStop = pos.getPosition() + readLength;
currentFeatureCount++;
}
} else {
if (currentFeatureChr != null) {
if (currentFeatureCount
> (minReadCount / MAX_READ_LENGTHS_PER_REGION) * minReadCandidateFraction) {
Feature region =
new Feature(
currentFeatureChr,
currentFeatureStart - readLength,
currentFeatureStop + readLength);
BreakpointCandidate candidate = new BreakpointCandidate(region, currentFeatureCount);
this.svCandidateRegions.add(candidate);
}
currentFeatureChr = null;
currentFeatureStart = -1;
currentFeatureStop = -1;
currentFeatureCount = 0;
} else {
currentFeatureChr = pos.getChromosome();
currentFeatureStart = pos.getPosition();
currentFeatureStop = pos.getPosition() + readLength;
currentFeatureCount = 1;
}
}
last = pos;
}
// Don't forget last SV candidate region
if (currentFeatureCount
> (minReadCount / MAX_READ_LENGTHS_PER_REGION) * minReadCandidateFraction) {
Feature region =
new Feature(
currentFeatureChr,
currentFeatureStart - readLength,
currentFeatureStop + readLength);
BreakpointCandidate candidate = new BreakpointCandidate(region, currentFeatureCount);
this.svCandidateRegions.add(candidate);
}
}
long end = System.currentTimeMillis();
int kmer = readLength + 1;
if (kmers.length > 0) {
kmer = kmers[0];
}
if (isDebug) {
System.err.println(
"Elapsed_msecs_in_NativeAssembler\tRegion:\t"
+ regions.get(0).getDescriptor()
+ "\tLength:\t"
+ regions.get(0).getLength()
+ "\tReadCount:\t"
+ readCount
+ "\tElapsed\t"
+ (end - start)
+ "\tAssembled\t"
+ isAssemblyCandidate
+ "\t"
+ kmer);
}
return contigs;
}
//
// Returns a downsampled set of reads for each sample.
//
private List<List<SAMRecord>> getReads(
List<String> inputFiles, List<Feature> regions, ReAligner realigner) {
int downsampleTarget = desiredNumberOfReads(regions);
List<DownsampledReadList> readsList = new ArrayList<DownsampledReadList>();
for (String input : inputFiles) {
Set<String> readIds = new HashSet<String>();
DownsampledReadList reads = new DownsampledReadList(downsampleTarget);
readsList.add(reads);
for (Feature region : regions) {
SAMFileReader reader = new SAMFileReader(new File(input));
reader.setValidationStringency(ValidationStringency.SILENT);
Iterator<SAMRecord> iter;
if (region != null) {
iter =
reader.queryOverlapping(
region.getSeqname(), (int) region.getStart(), (int) region.getEnd());
} else {
iter = reader.iterator();
}
while (iter.hasNext()) {
SAMRecord read = iter.next();
// Don't allow same read to be counted twice.
if ((!realigner.isFiltered(read))
&& (!read.getDuplicateReadFlag())
&& (!read.getReadFailsVendorQualityCheckFlag())
&& (read.getMappingQuality() >= realigner.getMinMappingQuality()
|| read.getReadUnmappedFlag())
&& (!readIds.contains(getIdentifier(read)))) {
if (read.getReadString().length() > readLength) {
reader.close();
throw new IllegalArgumentException(
"Maximum read length of: "
+ readLength
+ " exceeded for: "
+ read.getSAMString());
}
readIds.add(getIdentifier(read));
reads.add(read);
}
}
if (reads.getTotalReadCount() != reads.getReads().size()) {
if (isDebug) {
System.err.println(
"downsampled: "
+ regions.get(0).getDescriptor()
+ ": "
+ reads.getTotalReadCount()
+ " -> "
+ reads.getReads().size());
}
}
reader.close();
}
}
List<List<SAMRecord>> sampleReads = new ArrayList<List<SAMRecord>>();
for (DownsampledReadList downsampledReads : readsList) {
sampleReads.add(downsampledReads.getReads());
}
return sampleReads;
}
