@Test
public void testCountsFromLocusTraversal() {
final GenomeAnalysisEngine engine = new GenomeAnalysisEngine();
engine.setGenomeLocParser(genomeLocParser);
final Collection<SAMReaderID> samFiles = new ArrayList<>();
final SAMReaderID readerID = new SAMReaderID(testBAM, new Tags());
samFiles.add(readerID);
final SAMDataSource dataSource =
new SAMDataSource(
samFiles,
new ThreadAllocation(),
null,
genomeLocParser,
false,
SAMFileReader.ValidationStringency.STRICT,
null,
null,
new ValidationExclusion(),
new ArrayList<ReadFilter>(),
new ArrayList<ReadTransformer>(),
false,
(byte) 30,
false,
true);
engine.setReadsDataSource(dataSource);
final Set<String> samples = SampleUtils.getSAMFileSamples(dataSource.getHeader());
final TraverseLociNano traverseLociNano = new TraverseLociNano(1);
final DummyLocusWalker walker = new DummyLocusWalker();
traverseLociNano.initialize(engine, walker, null);
for (final Shard shard : dataSource.createShardIteratorOverAllReads(new LocusShardBalancer())) {
final WindowMaker windowMaker =
new WindowMaker(
shard, genomeLocParser, dataSource.seek(shard), shard.getGenomeLocs(), samples);
for (WindowMaker.WindowMakerIterator window : windowMaker) {
final LocusShardDataProvider dataProvider =
new LocusShardDataProvider(
shard,
shard.getReadProperties(),
genomeLocParser,
window.getLocus(),
window,
reference,
new ArrayList<ReferenceOrderedDataSource>());
traverseLociNano.traverse(walker, dataProvider, 0);
dataProvider.close();
}
windowMaker.close();
}
// dataSource.close();
Assert.assertEquals(
engine.getCumulativeMetrics().getNumReadsSeen(), contigs.size() * numReadsPerContig);
Assert.assertEquals(
engine.getCumulativeMetrics().getNumIterations(), contigs.size() * numReadsPerContig);
}
@Test
public void testFilteredCounts() {
final GenomeAnalysisEngine engine = new GenomeAnalysisEngine();
engine.setGenomeLocParser(genomeLocParser);
final Collection<SAMReaderID> samFiles = new ArrayList<>();
final SAMReaderID readerID = new SAMReaderID(testBAM, new Tags());
samFiles.add(readerID);
final List<ReadFilter> filters = new ArrayList<>();
filters.add(new EveryTenthReadFilter());
final SAMDataSource dataSource =
new SAMDataSource(
samFiles,
new ThreadAllocation(),
null,
genomeLocParser,
false,
SAMFileReader.ValidationStringency.STRICT,
null,
null,
new ValidationExclusion(),
filters,
new ArrayList<ReadTransformer>(),
false,
(byte) 30,
false,
true);
engine.setReadsDataSource(dataSource);
final TraverseReadsNano traverseReadsNano = new TraverseReadsNano(1);
final DummyReadWalker walker = new DummyReadWalker();
traverseReadsNano.initialize(engine, walker, null);
for (final Shard shard : dataSource.createShardIteratorOverAllReads(new ReadShardBalancer())) {
final ReadShardDataProvider dataProvider =
new ReadShardDataProvider(
shard,
engine.getGenomeLocParser(),
dataSource.seek(shard),
reference,
new ArrayList<ReferenceOrderedDataSource>());
traverseReadsNano.traverse(walker, dataProvider, 0);
dataProvider.close();
}
Assert.assertEquals(
(long)
engine
.getCumulativeMetrics()
.getCountsByFilter()
.get(EveryTenthReadFilter.class.getSimpleName()),
contigs.size() * numReadsPerContig / 10);
}
/**
* Makes association maps for the reads and loci coverage as described below :
*
* <p>- First: locusToReadMap -- a HashMap that describes for each locus, which reads contribute
* to its coverage. Note: Locus is in reference coordinates. Example: Locus => {read1, read2, ...,
* readN}
*
* <p>- Second: readToLocusMap -- a HashMap that describes for each read what loci it contributes
* to the coverage. Note: Locus is a boolean array, indexed from 0 (= startLocation) to N (=
* stopLocation), with value==true meaning it contributes to the coverage. Example: Read => {true,
* true, false, ... false}
*
* @param readList the list of reads to generate the association mappings
* @param startLocation the first reference coordinate of the region (inclusive)
* @param stopLocation the last reference coordinate of the region (inclusive)
* @return the two hashmaps described above
*/
public static Pair<HashMap<Integer, HashSet<GATKSAMRecord>>, HashMap<GATKSAMRecord, Boolean[]>>
getBothReadToLociMappings(List<GATKSAMRecord> readList, int startLocation, int stopLocation) {
int arraySize = stopLocation - startLocation + 1;
HashMap<Integer, HashSet<GATKSAMRecord>> locusToReadMap =
new HashMap<Integer, HashSet<GATKSAMRecord>>(2 * (stopLocation - startLocation + 1), 0.5f);
HashMap<GATKSAMRecord, Boolean[]> readToLocusMap =
new HashMap<GATKSAMRecord, Boolean[]>(2 * readList.size(), 0.5f);
for (int i = startLocation; i <= stopLocation; i++)
locusToReadMap.put(
i, new HashSet<GATKSAMRecord>()); // Initialize the locusToRead map with empty lists
for (GATKSAMRecord read : readList) {
readToLocusMap.put(
read, new Boolean[arraySize]); // Initialize the readToLocus map with empty arrays
int[] readCoverage = getCoverageDistributionOfRead(read, startLocation, stopLocation);
for (int i = 0; i < readCoverage.length; i++) {
int refLocation = i + startLocation;
if (readCoverage[i] > 0) {
// Update the hash for this locus
HashSet<GATKSAMRecord> readSet = locusToReadMap.get(refLocation);
readSet.add(read);
// Add this locus to the read hash
readToLocusMap.get(read)[refLocation - startLocation] = true;
} else
// Update the boolean array with a 'no coverage' from this read to this locus
readToLocusMap.get(read)[refLocation - startLocation] = false;
}
}
return new Pair<HashMap<Integer, HashSet<GATKSAMRecord>>, HashMap<GATKSAMRecord, Boolean[]>>(
locusToReadMap, readToLocusMap);
}
public static void align(
Graph graph,
SAMRecord rec,
Node recNode,
ReferenceSequence sequence,
SAMProgramRecord programRecord,
int offset,
AlleleCoverageCutoffs alleleCoverageCutoffs,
boolean correctBases,
boolean useSequenceQualities,
int MAXIMUM_TOTAL_COVERAGE,
int MAX_HEAP_SIZE)
throws Exception {
int i;
AlignHeapNode curAlignHeapNode = null;
AlignHeapNode nextAlignHeapNode = null;
AlignHeapNode bestAlignHeapNode = null;
AlignHeap heap = null;
String read = null; // could be cs
String readBases = null; // always nt
String qualities = null; // could be cq
SRMAUtil.Space space = SRMAUtil.Space.NTSPACE;
ListIterator<NodeRecord> iter = null;
AlignHeapNodeComparator comp = null;
int alignmentStart = -1;
int numStartNodesAdded = 0;
boolean strand = rec.getReadNegativeStrandFlag(); // false -> forward, true -> reverse
String softClipStartBases = null;
String softClipStartQualities = null;
String softClipEndBases = null;
String softClipEndQualities = null;
// Debugging stuff
String readName = rec.getReadName();
assert SRMAUtil.Space.COLORSPACE != space;
// Get space
read = (String) rec.getAttribute("CS");
if (null == read) {
// Use base space
space = SRMAUtil.Space.NTSPACE;
} else {
// assumes CS and CQ are always in sequencing order
space = SRMAUtil.Space.COLORSPACE;
}
// Get read and qualities
if (space == SRMAUtil.Space.NTSPACE) {
byte tmpRead[] = rec.getReadString().getBytes();
byte tmpQualities[] = rec.getBaseQualityString().getBytes();
// Reverse once
if (strand) { // reverse
SAMRecordUtil.reverseArray(tmpRead);
SAMRecordUtil.reverseArray(tmpQualities);
}
read = new String(tmpRead);
readBases = new String(tmpRead);
qualities = new String(tmpQualities);
// Reverse again
if (strand) { // reverse
SAMRecordUtil.reverseArray(tmpRead);
SAMRecordUtil.reverseArray(tmpQualities);
}
} else {
byte tmpRead[] = rec.getReadString().getBytes();
// Reverse once
if (strand) { // reverse
SAMRecordUtil.reverseArray(tmpRead);
}
readBases = new String(tmpRead);
// Reverse again
if (strand) { // reverse
SAMRecordUtil.reverseArray(tmpRead);
}
read = SRMAUtil.normalizeColorSpaceRead(read);
qualities = (String) rec.getAttribute("CQ");
// Some aligners include a quality value for the adapter. A quality value
// IMHO should not be given for an unobserved (assumed) peice of data. Trim
// the first quality in this case
if (qualities.length() == 1 + read.length()) { // trim the first quality
qualities = qualities.substring(1);
}
}
// Reverse back
if (readBases.length() <= 0) {
throw new Exception("Error. The current alignment has no bases.");
}
if (read.length() <= 0) {
throw new Exception("Error. The current alignment has no bases.");
}
if (qualities.length() <= 0) {
throw new Exception("Error. The current alignment has no qualities.");
}
if (readBases.length() != read.length()) {
if (space == SRMAUtil.Space.COLORSPACE) {
throw new Exception(
"Error. The current alignment's read bases length does not match the length of the colors in the CS tag ["
+ rec.getReadName()
+ "].");
} else {
throw new Exception("Error. Internal error: readBases.length() != read.length()");
}
}
// Deal with soft-clipping
// - save the soft clipped sequence for latter
{
List<CigarElement> cigarElements = null;
cigarElements = rec.getCigar().getCigarElements();
CigarElement e1 = cigarElements.get(0); // first
CigarElement e2 = cigarElements.get(cigarElements.size() - 1); // last
// Soft-clipped
if (CigarOperator.S == e1.getOperator()) {
if (space == SRMAUtil.Space.COLORSPACE) {
throw new Exception(
"Error. Soft clipping with color-space data not currently supported.");
}
int l = e1.getLength();
if (strand) { // reverse
softClipStartBases = readBases.substring(readBases.length() - l);
softClipStartQualities = qualities.substring(qualities.length() - l);
readBases = readBases.substring(0, readBases.length() - l);
read = read.substring(0, read.length() - l);
qualities = qualities.substring(0, qualities.length() - l);
} else {
softClipStartBases = readBases.substring(0, l - 1);
softClipStartQualities = qualities.substring(0, l - 1);
readBases = readBases.substring(l);
read = read.substring(l);
qualities = qualities.substring(l);
}
}
if (CigarOperator.S == e2.getOperator()) {
if (space == SRMAUtil.Space.COLORSPACE) {
throw new Exception(
"Error. Soft clipping with color-space data not currently supported.");
}
int l = e2.getLength();
if (strand) { // reverse
softClipEndBases = readBases.substring(0, l - 1);
softClipEndQualities = qualities.substring(0, l - 1);
readBases = readBases.substring(l);
read = read.substring(l);
qualities = qualities.substring(l);
} else {
softClipEndBases = readBases.substring(readBases.length() - l);
softClipEndQualities = qualities.substring(qualities.length() - l);
readBases = readBases.substring(0, readBases.length() - l);
read = read.substring(0, read.length() - l);
qualities = qualities.substring(0, qualities.length() - l);
}
}
}
// Remove mate pair information
Align.removeMateInfo(rec);
comp =
new AlignHeapNodeComparator(
(strand) ? AlignHeap.HeapType.MAXHEAP : AlignHeap.HeapType.MINHEAP);
// Bound by original alignment if possible
bestAlignHeapNode =
Align.boundWithOriginalAlignment(
rec,
graph,
recNode,
comp,
strand,
read,
qualities,
readBases,
space,
sequence,
alleleCoverageCutoffs,
useSequenceQualities,
MAXIMUM_TOTAL_COVERAGE,
MAX_HEAP_SIZE);
/*
System.err.println("readName="+rec.getReadName());
if(null != bestAlignHeapNode) {
System.err.println("\nFOUND BEST:" + rec.toString());
}
else {
System.err.println("\nNOT FOUND (BEST): " + rec.toString());
}
Align.updateSAM(rec, programRecord, bestAlignHeapNode, space, read, qualities, softClipStartBases, softClipStartQualities, softClipEndBases, softClipEndQualities, strand, correctBases);
return;
*/
heap = new AlignHeap((strand) ? AlignHeap.HeapType.MAXHEAP : AlignHeap.HeapType.MINHEAP);
// Add start nodes
if (strand) { // reverse
alignmentStart = rec.getAlignmentEnd();
for (i = alignmentStart + offset; alignmentStart - offset <= i; i--) {
int position = graph.getPriorityQueueIndexAtPositionOrBefore(i);
PriorityQueue<Node> startNodeQueue = graph.getPriorityQueue(position);
if (0 != position && null != startNodeQueue) {
Iterator<Node> startNodeQueueIter = startNodeQueue.iterator();
while (startNodeQueueIter.hasNext()) {
Node startNode = startNodeQueueIter.next();
int f = passFilters(graph, startNode, alleleCoverageCutoffs, MAXIMUM_TOTAL_COVERAGE);
if (0 == f) {
heap.add(
new AlignHeapNode(
null,
startNode,
startNode.coverage,
read.charAt(0),
qualities.charAt(0),
useSequenceQualities,
space));
} else if (f < 0) {
return;
}
if (startNode.position < i) {
i = startNode.position;
}
numStartNodesAdded++;
}
}
}
} else {
alignmentStart = rec.getAlignmentStart();
for (i = alignmentStart - offset; i <= alignmentStart + offset; i++) {
int position = graph.getPriorityQueueIndexAtPositionOrGreater(i);
PriorityQueue<Node> startNodeQueue = graph.getPriorityQueue(position);
if (0 != position && null != startNodeQueue) {
Iterator<Node> startNodeQueueIter = startNodeQueue.iterator();
while (startNodeQueueIter.hasNext()) {
Node startNode = startNodeQueueIter.next();
int f = passFilters(graph, startNode, alleleCoverageCutoffs, MAXIMUM_TOTAL_COVERAGE);
if (0 == f) {
heap.add(
new AlignHeapNode(
null,
startNode,
startNode.coverage,
read.charAt(0),
qualities.charAt(0),
useSequenceQualities,
space));
} else if (f < 0) {
return;
}
if (i < startNode.position) {
i = startNode.position;
}
numStartNodesAdded++;
}
}
}
}
if (numStartNodesAdded == 0) {
throw new Exception("Did not add any start nodes!");
}
// Get first node off the heap
curAlignHeapNode = heap.poll();
while (null != curAlignHeapNode) {
if (MAX_HEAP_SIZE <= heap.size()) {
// too many to consider
return;
}
// System.err.println("strand:" + strand + "\tsize:" + heap.size() + "\talignmentStart:" +
// alignmentStart + "\toffset:" + offset + "\treadOffset:" + curAlignHeapNode.readOffset);
// System.err.print("size:" + heap.size() + ":" + curAlignHeapNode.readOffset + ":" +
// curAlignHeapNode.score + ":" + curAlignHeapNode.alleleCoverageSum + ":" +
// curAlignHeapNode.startPosition + "\t");
// curAlignHeapNode.node.print(System.err);
// System.err.print("\rposition:" + curAlignHeapNode.node.position + "\treadOffset:" +
// curAlignHeapNode.readOffset);
// Remove all non-insertions with the same contig/pos/read-offset/type/base and lower score
nextAlignHeapNode = heap.peek();
while (Node.INSERTION != curAlignHeapNode.node.type
&& null != nextAlignHeapNode
&& 0 == comp.compare(curAlignHeapNode, nextAlignHeapNode)) {
if (curAlignHeapNode.score < nextAlignHeapNode.score
|| (curAlignHeapNode.score == nextAlignHeapNode.score
&& curAlignHeapNode.alleleCoverageSum < nextAlignHeapNode.alleleCoverageSum)) {
// Update current node
curAlignHeapNode = heap.poll();
} else {
// Ignore next node
heap.poll();
}
nextAlignHeapNode = heap.peek();
}
nextAlignHeapNode = null;
// Check if the alignment is complete
if (curAlignHeapNode.readOffset == read.length() - 1) {
// All read bases examined, store if has the best alignment.
// System.err.print(curAlignHeapNode.alleleCoverageSum + ":" + curAlignHeapNode.score +
// ":");
// System.err.print(curAlignHeapNode.startPosition + ":");
// curAlignHeapNode.node.print(System.err);
if (null == bestAlignHeapNode
|| bestAlignHeapNode.score < curAlignHeapNode.score
|| (bestAlignHeapNode.score == curAlignHeapNode.score
&& bestAlignHeapNode.alleleCoverageSum < curAlignHeapNode.alleleCoverageSum)) {
bestAlignHeapNode = curAlignHeapNode;
}
} else if (null != bestAlignHeapNode && curAlignHeapNode.score < bestAlignHeapNode.score) {
// ignore, under the assumption that scores can only become more negative.
} else {
if (strand) { // reverse
// Go to all the "prev" nodes
iter = curAlignHeapNode.node.prev.listIterator();
} else { // forward
// Go to all "next" nodes
iter = curAlignHeapNode.node.next.listIterator();
}
while (iter.hasNext()) {
NodeRecord next = iter.next();
int f =
passFilters(
graph, next.node, next.coverage, alleleCoverageCutoffs, MAXIMUM_TOTAL_COVERAGE);
if (0 == f) {
heap.add(
new AlignHeapNode(
curAlignHeapNode,
next.node,
next.coverage,
read.charAt(curAlignHeapNode.readOffset + 1),
qualities.charAt(curAlignHeapNode.readOffset + 1),
useSequenceQualities,
space));
} else if (f < 0) {
return;
}
}
iter = null;
}
// Get next node
curAlignHeapNode = heap.poll();
}
// Recover alignment
Align.updateSAM(
rec,
sequence,
programRecord,
bestAlignHeapNode,
space,
read,
qualities,
softClipStartBases,
softClipStartQualities,
softClipEndBases,
softClipEndQualities,
strand,
correctBases);
}