@Override
protected void doWork(String inputSource, VcfIterator r, VariantContextWriter w)
throws IOException {
VCFHeader header = r.getHeader();
VCFHeader h2 = new VCFHeader(header.getMetaDataInInputOrder(), header.getSampleNamesInOrder());
h2.addMetaDataLine(
new VCFInfoHeaderLine(
TAG,
VCFHeaderLineCount.UNBOUNDED,
VCFHeaderLineType.String,
"metadata added from " + TABIX + " . Format was " + FORMAT));
h2.addMetaDataLine(
new VCFHeaderLine(
getClass().getSimpleName() + "CmdLine", String.valueOf(getProgramCommandLine())));
h2.addMetaDataLine(
new VCFHeaderLine(getClass().getSimpleName() + "Version", String.valueOf(getVersion())));
h2.addMetaDataLine(
new VCFHeaderLine(
getClass().getSimpleName() + "HtsJdkVersion", HtsjdkVersion.getVersion()));
h2.addMetaDataLine(
new VCFHeaderLine(getClass().getSimpleName() + "HtsJdkHome", HtsjdkVersion.getHome()));
SAMSequenceDictionaryProgress progress = new SAMSequenceDictionaryProgress(header);
w.writeHeader(h2);
while (r.hasNext()) {
VariantContext ctx = progress.watch(r.next());
Set<String> annotations = new HashSet<String>();
CloseableIterator<BedLine> iter =
this.bedReader.iterator(ctx.getContig(), ctx.getStart() - 1, ctx.getEnd() + 1);
while (iter.hasNext()) {
BedLine bedLine = iter.next();
if (!ctx.getContig().equals(bedLine.getContig())) continue;
if (ctx.getStart() - 1 >= bedLine.getEnd()) continue;
if (ctx.getEnd() - 1 < bedLine.getStart()) continue;
String newannot = this.parsedFormat.toString(bedLine);
if (!newannot.isEmpty()) annotations.add(VCFUtils.escapeInfoField(newannot));
}
CloserUtil.close(iter);
if (annotations.isEmpty()) {
w.add(ctx);
continue;
}
VariantContextBuilder vcb = new VariantContextBuilder(ctx);
vcb.attribute(TAG, annotations.toArray());
w.add(vcb.make());
incrVariantCount();
if (checkOutputError()) break;
}
progress.finish();
}
@Override
public void accumulate(final VariantContext ctx) {
logger.record(ctx.getContig(), ctx.getStart());
final String variantChrom = ctx.getContig();
final int variantPos = ctx.getStart();
// Skip anything a little too funky
if (ctx.isFiltered()) return;
if (!ctx.isVariant()) return;
if (SKIP_CHROMS.contains(variantChrom)) return;
for (final MendelianViolationMetrics trio : trios) {
final Genotype momGt = ctx.getGenotype(trio.MOTHER);
final Genotype dadGt = ctx.getGenotype(trio.FATHER);
final Genotype kidGt = ctx.getGenotype(trio.OFFSPRING);
// if any genotype:
// - has a non-snp allele; or
// - lacks a reference allele
//
// then ignore this trio
if (CollectionUtil.makeList(momGt, dadGt, kidGt)
.stream()
.anyMatch(
gt ->
gt.isHetNonRef()
|| Stream.concat(Stream.of(ctx.getReference()), gt.getAlleles().stream())
.anyMatch(a -> a.length() != 1 || a.isSymbolic()))) {
continue;
}
// if between the trio there are more than 2 alleles including the reference, continue
if (Stream.concat(
Collections.singleton(ctx.getReference()).stream(),
CollectionUtil.makeList(momGt, dadGt, kidGt)
.stream()
.flatMap(gt -> gt.getAlleles().stream()))
.collect(Collectors.toSet())
.size()
> 2) continue;
// Test to make sure:
// 1) That the site is in fact variant in the trio
// 2) that the offspring doesn't have a really wacky het allele balance
if (!isVariant(momGt, dadGt, kidGt)) continue;
if (kidGt.isHet()) {
final int[] ad = kidGt.getAD();
if (ad == null) continue;
final List<Integer> adOfAlleles =
kidGt
.getAlleles()
.stream()
.map(a -> ad[ctx.getAlleleIndex(a)])
.collect(Collectors.toList());
final double minAlleleFraction =
Math.min(adOfAlleles.get(0), adOfAlleles.get(1))
/ (double) (adOfAlleles.get(0) + adOfAlleles.get(1));
if (minAlleleFraction < MIN_HET_FRACTION) continue;
}
///////////////////////////////////////////////////////////////
// Determine whether the offspring should be haploid at this
// locus and which is the parental donor of the haploid genotype
///////////////////////////////////////////////////////////////
boolean haploid = false;
Genotype haploidParentalGenotype = null;
if (FEMALE_CHROMS.contains(variantChrom) && trio.OFFSPRING_SEX != Sex.Unknown) {
if (trio.OFFSPRING_SEX == Sex.Female) {
// famale
haploid = false;
} else if (isInPseudoAutosomalRegion(variantChrom, variantPos)) {
// male but in PAR on X, so diploid
haploid = false;
} else {
// male, out of PAR on X, haploid
haploid = true;
haploidParentalGenotype = momGt;
}
}
// the PAR on the male chromosome should be masked so that reads
// align to the female chromosomes instead, so there's no point
// of worrying about that here.
if (MALE_CHROMS.contains(variantChrom)) {
if (trio.OFFSPRING_SEX == Sex.Male) {
haploid = true;
haploidParentalGenotype = dadGt;
} else {
continue;
}
}
// We only want to look at sites where we have high enough confidence that the genotypes we
// are looking at are
// interesting. We want to ensure that parents are always GQ>=MIN_GQ, and that the kid is
// either GQ>=MIN_GQ or in the
// case where kid is het that the phred-scaled-likelihood of being reference is >=MIN_GQ.
if (haploid
&& (haploidParentalGenotype.isNoCall() || haploidParentalGenotype.getGQ() < MIN_GQ))
continue;
if (!haploid
&& (momGt.isNoCall()
|| momGt.getGQ() < MIN_GQ
|| dadGt.isNoCall()
|| dadGt.getGQ() < MIN_GQ)) continue;
if (kidGt.isNoCall()) continue;
if (momGt.isHomRef() && dadGt.isHomRef() && !kidGt.isHomRef()) {
if (kidGt.getPL()[0] < MIN_GQ) continue;
} else if (kidGt.getGQ() < MIN_GQ) continue;
// Also filter on the DP for each of the samples - it's possible to miss hets when DP is too
// low
if (haploid && (kidGt.getDP() < MIN_DP || haploidParentalGenotype.getDP() < MIN_DP)) continue;
if (!haploid && (kidGt.getDP() < MIN_DP || momGt.getDP() < MIN_DP || dadGt.getDP() < MIN_DP))
continue;
trio.NUM_VARIANT_SITES++;
///////////////////////////////////////////////////////////////
// First test for haploid violations
///////////////////////////////////////////////////////////////
MendelianViolation type = null;
if (haploid) {
if (kidGt.isHet()) continue; // Should not see heterozygous calls at haploid regions
if (!haploidParentalGenotype.getAlleles().contains(kidGt.getAllele(0))) {
if (kidGt.isHomRef()) {
type = MendelianViolation.Haploid_Other;
trio.NUM_HAPLOID_OTHER++;
} else {
type = MendelianViolation.Haploid_Denovo;
trio.NUM_HAPLOID_DENOVO++;
}
}
}
///////////////////////////////////////////////////////////////
// Then test for diploid mendelian violations
///////////////////////////////////////////////////////////////
else if (isMendelianViolation(momGt, dadGt, kidGt)) {
if (momGt.isHomRef() && dadGt.isHomRef() && !kidGt.isHomRef()) {
trio.NUM_DIPLOID_DENOVO++;
type = MendelianViolation.Diploid_Denovo;
} else if (momGt.isHomVar() && dadGt.isHomVar() && kidGt.isHet()) {
trio.NUM_HOMVAR_HOMVAR_HET++;
type = MendelianViolation.HomVar_HomVar_Het;
} else if (kidGt.isHom()
&& ((momGt.isHomRef() && dadGt.isHomVar()) || (momGt.isHomVar() && dadGt.isHomRef()))) {
trio.NUM_HOMREF_HOMVAR_HOM++;
type = MendelianViolation.HomRef_HomVar_Hom;
} else if (kidGt.isHom()
&& ((momGt.isHom() && dadGt.isHet()) || (momGt.isHet() && dadGt.isHom()))) {
trio.NUM_HOM_HET_HOM++;
type = MendelianViolation.Hom_Het_Hom;
} else {
trio.NUM_OTHER++;
type = MendelianViolation.Other;
}
}
// Output a record into the family's violation VCF
if (type != null) {
// Create a new Context subsetted to the three samples
final VariantContextBuilder builder = new VariantContextBuilder(ctx);
builder.genotypes(
ctx.getGenotypes()
.subsetToSamples(CollectionUtil.makeSet(trio.MOTHER, trio.FATHER, trio.OFFSPRING)));
builder.attribute(MENDELIAN_VIOLATION_KEY, type.name());
// Copy over some useful attributes from the full context
if (ctx.hasAttribute(VCFConstants.ALLELE_COUNT_KEY))
builder.attribute(ORIGINAL_AC, ctx.getAttribute(VCFConstants.ALLELE_COUNT_KEY));
if (ctx.hasAttribute(VCFConstants.ALLELE_FREQUENCY_KEY))
builder.attribute(ORIGINAL_AF, ctx.getAttribute(VCFConstants.ALLELE_FREQUENCY_KEY));
if (ctx.hasAttribute(VCFConstants.ALLELE_NUMBER_KEY))
builder.attribute(ORIGINAL_AN, ctx.getAttribute(VCFConstants.ALLELE_NUMBER_KEY));
// Write out the variant record
familyToViolations.get(trio.FAMILY_ID).add(builder.make());
}
}
}
@Override
protected Object doWork() {
IOUtil.assertFileIsReadable(INPUT);
IOUtil.assertFileIsReadable(REFERENCE_SEQUENCE);
IOUtil.assertFileIsReadable(CHAIN);
IOUtil.assertFileIsWritable(OUTPUT);
IOUtil.assertFileIsWritable(REJECT);
////////////////////////////////////////////////////////////////////////
// Setup the inputs
////////////////////////////////////////////////////////////////////////
final LiftOver liftOver = new LiftOver(CHAIN);
final VCFFileReader in = new VCFFileReader(INPUT, false);
logger.info("Loading up the target reference genome.");
final ReferenceSequenceFileWalker walker = new ReferenceSequenceFileWalker(REFERENCE_SEQUENCE);
final Map<String, byte[]> refSeqs = new HashMap<>();
for (final SAMSequenceRecord rec : walker.getSequenceDictionary().getSequences()) {
refSeqs.put(rec.getSequenceName(), walker.get(rec.getSequenceIndex()).getBases());
}
CloserUtil.close(walker);
////////////////////////////////////////////////////////////////////////
// Setup the outputs
////////////////////////////////////////////////////////////////////////
final VCFHeader inHeader = in.getFileHeader();
final VCFHeader outHeader = new VCFHeader(inHeader);
outHeader.setSequenceDictionary(walker.getSequenceDictionary());
final VariantContextWriter out =
new VariantContextWriterBuilder()
.setOption(Options.INDEX_ON_THE_FLY)
.setOutputFile(OUTPUT)
.setReferenceDictionary(walker.getSequenceDictionary())
.build();
out.writeHeader(outHeader);
final VariantContextWriter rejects =
new VariantContextWriterBuilder()
.setOutputFile(REJECT)
.unsetOption(Options.INDEX_ON_THE_FLY)
.build();
final VCFHeader rejectHeader = new VCFHeader(in.getFileHeader());
for (final VCFFilterHeaderLine line : FILTERS) rejectHeader.addMetaDataLine(line);
rejects.writeHeader(rejectHeader);
////////////////////////////////////////////////////////////////////////
// Read the input VCF, lift the records over and write to the sorting
// collection.
////////////////////////////////////////////////////////////////////////
long failedLiftover = 0, failedAlleleCheck = 0, total = 0;
logger.info("Lifting variants over and sorting.");
final SortingCollection<VariantContext> sorter =
SortingCollection.newInstance(
VariantContext.class,
new VCFRecordCodec(outHeader),
outHeader.getVCFRecordComparator(),
MAX_RECORDS_IN_RAM,
TMP_DIR);
ProgressLogger progress = new ProgressLogger(logger, 1000000, "read");
for (final VariantContext ctx : in) {
++total;
final Interval source =
new Interval(
ctx.getContig(),
ctx.getStart(),
ctx.getEnd(),
false,
ctx.getContig() + ":" + ctx.getStart() + "-" + ctx.getEnd());
final Interval target = liftOver.liftOver(source, 1.0);
if (target == null) {
rejects.add(new VariantContextBuilder(ctx).filter(FILTER_CANNOT_LIFTOVER).make());
failedLiftover++;
} else {
// Fix the alleles if we went from positive to negative strand
final List<Allele> alleles = new ArrayList<>();
for (final Allele oldAllele : ctx.getAlleles()) {
if (target.isPositiveStrand() || oldAllele.isSymbolic()) {
alleles.add(oldAllele);
} else {
alleles.add(
Allele.create(
SequenceUtil.reverseComplement(oldAllele.getBaseString()),
oldAllele.isReference()));
}
}
// Build the new variant context
final VariantContextBuilder builder =
new VariantContextBuilder(
ctx.getSource(), target.getContig(), target.getStart(), target.getEnd(), alleles);
builder.id(ctx.getID());
builder.attributes(ctx.getAttributes());
builder.genotypes(ctx.getGenotypes());
builder.filters(ctx.getFilters());
builder.log10PError(ctx.getLog10PError());
// Check that the reference allele still agrees with the reference sequence
boolean mismatchesReference = false;
for (final Allele allele : builder.getAlleles()) {
if (allele.isReference()) {
final byte[] ref = refSeqs.get(target.getContig());
final String refString =
StringUtil.bytesToString(ref, target.getStart() - 1, target.length());
if (!refString.equalsIgnoreCase(allele.getBaseString())) {
mismatchesReference = true;
}
break;
}
}
if (mismatchesReference) {
rejects.add(new VariantContextBuilder(ctx).filter(FILTER_MISMATCHING_REF_ALLELE).make());
failedAlleleCheck++;
} else {
sorter.add(builder.make());
}
}
progress.record(ctx.getContig(), ctx.getStart());
}
final NumberFormat pfmt = new DecimalFormat("0.0000%");
final String pct = pfmt.format((failedLiftover + failedAlleleCheck) / (double) total);
logger.info("Processed ", total, " variants.");
logger.info(Long.toString(failedLiftover), " variants failed to liftover.");
logger.info(
Long.toString(failedAlleleCheck),
" variants lifted over but had mismatching reference alleles after lift over.");
logger.info(pct, " of variants were not successfully lifted over and written to the output.");
rejects.close();
in.close();
////////////////////////////////////////////////////////////////////////
// Write the sorted outputs to the final output file
////////////////////////////////////////////////////////////////////////
sorter.doneAdding();
progress = new ProgressLogger(logger, 1000000, "written");
logger.info("Writing out sorted records to final VCF.");
for (final VariantContext ctx : sorter) {
out.add(ctx);
progress.record(ctx.getContig(), ctx.getStart());
}
out.close();
sorter.cleanup();
return null;
}
@Override
public void runCommand() {
logger.info("MergeVCFColumnsCommand");
/*
* Assumptions
* (1) Only two vcfs that are sorted with the same contig order
* (2) if contigs on it same order, then we will just skip that contig
* (3) No overlapping samples allowed
*
* Output:
* A vcf where intersecting sites are merged together and will only return biallelic markers
* the info field will be cleared
* the only GT FORMAT field will be there
*/
Collection<File> vcfs = applicationOptions.getVcfs();
String outfile = applicationOptions.getOutFile();
if (vcfs.size() != 2) {
throw new IllegalArgumentException("This function requires exactly two vcfs");
}
Iterator<File> vcfFileIter = vcfs.iterator();
File vcf1 = vcfFileIter.next();
File vcf2 = vcfFileIter.next();
VCFFileReader reader1 = new VCFFileReader(vcf1, false);
VCFFileReader reader2 = new VCFFileReader(vcf2, false);
Iterator<VariantContext> iter1 = reader1.iterator();
Iterator<VariantContext> iter2 = reader2.iterator();
VariantContextComparator comparator = new VariantContextComparator();
/*
* Merge headers
*/
VCFHeader header1 = reader1.getFileHeader();
VCFHeader header2 = reader2.getFileHeader();
List<String> samples1 = header1.getGenotypeSamples();
List<String> samples2 = header2.getGenotypeSamples();
List<String> mergedSamples = new ArrayList<>(samples1.size() + samples2.size());
mergedSamples.addAll(samples1);
mergedSamples.addAll(samples2);
// Validate that there are no duplicates
HashSet<String> sampleSet = new HashSet<String>();
for (String id : mergedSamples) {
if (sampleSet.contains(id)) {
throw new IllegalArgumentException("Duplicate id found: " + id);
} else {
sampleSet.add(id);
}
}
HashSet<VCFHeaderLine> meta = new HashSet<>();
meta.add(new VCFFormatHeaderLine("GT", 1, VCFHeaderLineType.String, "GT"));
meta.addAll(header1.getContigLines());
VCFHeader mergedHeader = new VCFHeader(meta, mergedSamples);
/*
* Create encoder
*/
VCFEncoder encoder = new VCFEncoder(mergedHeader, false, false);
BufferedWriter writer = null;
try {
if (outfile.endsWith(".gz")) {
BlockCompressedOutputStream outstream = new BlockCompressedOutputStream(new File(outfile));
writer = new BufferedWriter(new OutputStreamWriter(outstream));
} else {
writer =
Files.newBufferedWriter(
Paths.get(outfile),
Charset.defaultCharset(),
StandardOpenOption.CREATE,
StandardOpenOption.TRUNCATE_EXISTING);
}
/*
* Write header
*/
VCFHeaderWriter.writeHeader(writer, mergedHeader);
logger.info("Wrote header");
VariantContext previous1 = null;
VariantContext previous2 = null;
int count = 0;
int countFile1 = 0;
int countFile2 = 0;
boolean usePrevious1 = false;
boolean usePrevious2 = false;
while (iter1.hasNext() || iter2.hasNext()) {
if ((iter1.hasNext() || usePrevious1) && (iter2.hasNext() || usePrevious2)) {
VariantContext variant1 = null;
VariantContext variant2 = null;
// if(usePrevious1 == true && usePrevious2 == true &&
// comparator.compare(previous1,previous2) != 0) {
// //then skip both
// usePrevious1 = false;
// usePrevious2 = false;
// }
if (usePrevious1) {
variant1 = previous1;
} else {
variant1 = iter1.next();
countFile1++;
}
if (usePrevious2) {
variant2 = previous2;
} else {
variant2 = iter2.next();
countFile2++;
}
// check that variants are ordered correctly
if (previous1 != null
&& previous1 != variant1
&& comparator.compare(previous1, variant1) > 0) {
throw new IllegalStateException(
previous1.getContig()
+ ":"
+ previous1.getStart()
+ " > "
+ variant1.getContig()
+ ":"
+ variant1.getStart());
}
if (previous2 != null
&& previous2 != variant2
&& comparator.compare(previous2, variant2) > 0) {
throw new IllegalStateException(
previous2.getContig()
+ ":"
+ previous2.getStart()
+ " > "
+ variant2.getContig()
+ ":"
+ variant2.getStart());
}
int cmp = comparator.compare(variant1, variant2);
if (cmp < 0) {
// logger.info("Skipping VCF1: " + variant1.getContig() + ":" + variant1.getStart() +
// "\t" + variant1.getReference().toString() + "\t" + variant1.getAlternateAlleles());
if (usePrevious1 == true && usePrevious2 == true) {
// variant1 < variant2
// we need to go to next variant in vcf1
usePrevious1 = false;
}
usePrevious2 = true;
} else if (cmp > 0) {
if (usePrevious1 == true && usePrevious2 == true) {
// variant1 > variant2
// we need to go to next variant in vcf2
usePrevious2 = false;
}
usePrevious1 = true;
// logger.info("Skipping VCF2: " + variant2.getContig() + ":" + variant2.getStart() +
// "\t" + variant2.getReference().toString() + "\t" + variant2.getAlternateAlleles());
} else {
// they equal position
usePrevious1 = false;
usePrevious2 = false;
if (variant1.isBiallelic()
&& variant2.isBiallelic()
&& variant1.getReference().equals(variant2.getReference())
&& variant1.getAlternateAllele(0).equals(variant2.getAlternateAllele(0))) {
// TODO: Finish merging
// both variants are bialleleic and the reference and alternative alleles match
count++;
if (count % 10000 == 0) {
logger.info(count + " mergeable variants found");
}
VariantContext merged = VariantContextMerger.merge(variant1, variant2);
writer.write(encoder.encode(merged));
writer.write("\n");
} else {
// skip if they do not equal
// logger.info("Skipping: " + variant1.getContig() + ":" + variant1.getStart() +
// "\t" + variant1.getReference().toString() + "\t" +
// variant1.getAlternateAlleles());
// logger.info("Skipping: " + variant2.getContig() + ":" + variant2.getStart() +
// "\t" + variant2.getReference().toString() + "\t" +
// variant2.getAlternateAlleles());
}
}
previous1 = variant1;
previous2 = variant2;
} else if (iter1.hasNext()) {
// just skip remaining variants
VariantContext current = iter1.next();
countFile1++;
if (previous1 != null && current != null && comparator.compare(previous1, current) > 0) {
throw new IllegalStateException(
previous1.getContig()
+ ":"
+ previous1.getStart()
+ " > "
+ current.getContig()
+ ":"
+ current.getStart());
}
previous1 = current;
// logger.info("Skipping: " + previous1.getContig() + ":" + previous1.getStart() + "\t" +
// previous1.getReference().toString() + "\t" + previous1.getAlternateAlleles());
} else if (iter2.hasNext()) {
// just skip remaining variants
// fixed bug/ was iter1 changed to iter2
VariantContext current = iter2.next();
countFile2++;
if (previous2 != null && current != null && comparator.compare(previous2, current) > 0) {
throw new IllegalStateException(
previous2.getContig()
+ ":"
+ previous2.getStart()
+ " > "
+ current.getContig()
+ ":"
+ current.getStart());
}
previous2 = current;
// logger.info("Skipping: " + previous2.getContig() + ":" + previous2.getStart() + "\t" +
// previous2.getReference().toString() + "\t" + previous2.getAlternateAlleles());
} else {
throw new IllegalStateException("Error should not of reached this point");
}
}
reader1.close();
reader2.close();
logger.info(count + " merged variants");
logger.info(countFile1 + " variants in " + vcf1.getAbsolutePath());
logger.info(countFile2 + " variants in " + vcf2.getAbsolutePath());
} catch (Exception e) {
e.printStackTrace();
} finally {
if (writer != null) {
try {
logger.info("Flushing writer");
writer.close();
} catch (IOException e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
}
}
logger.info("finished merging vcfs");
}
