broadinstitute/gatk 4.1.5.0 on GitHub

Download release: gatk-4.1.5.0.zip
Docker image: https://hub.docker.com/r/broadinstitute/gatk/

Highlights of the 4.1.5.0 release:

A new, improved version of the --linked-de-bruijn-graph mode for HaplotypeCaller and Mutect2 that has better sensitivity compared to the previous linked DeBruijn graph implementation (#6394)
A new version of GenomicsDB that fixes many frequently-reported issues
LeftAlignIndels now works for multiple indels
VariantAnnotator and Concordance are now out of beta
A significant number of bug fixes to major tools like GenotypeGVCFs and SelectVariants

Full list of changes:

HaplotypeCaller
- New, improved version of the --linked-de-bruijn-graph mode for HaplotypeCaller and Mutect2 that has better sensitivity compared to the previous linked DeBruijn graph implementation (#6394)
  - Running HaplotypeCaller in this mode will reduce the number of erroneous haplotypes discovered which can improve genotyping, phasing, and runtime.
  - Changed the haplotype recovery step to check that it covers all paths through the graph even if there are poorly supported paths in the JunctionTrees. Added the argument --disable-artificial-haplotype-recovery to disable this behavior.
  - Added the ability to expand graph kmer size after haplotype recovery in the event that there was a failure due to overcomplicated assembly graphs.
  - Added code to squeeze extra sensitivity out of the junction trees by tolerating SNP errors when threading the junction trees themselves
- Realigning to best haplotype handles indels better (#6461)
- Fixed issue #5434 on inconsistent selection of reads for the PL, AD, and DP calculations. (#6055)
- Fixed bug where SNP and indel pseudocounts were swapped in the AlleleFrequencyCalculator (#6401)
- The qual used in HaplotypeCaller's isActive() method now matches that of GenotypeGVCFs. That is, they both now use the new qual. (#6343)
- Skip non-nucleotide alleles in force-calling mode, fixing bug (#6405)
- Fixed the hidden/experimental --error-correct-reads argument to actually correct the bases and qualities (#6366)
- Removed the deprecated and obsolete --use-new-qual-calculator argument (#6398)
- Refactored code related to windows and padding for assembly and genotyping, with slight changes to HMM padding for indels (#6358)
Mutect2
- Improved SomaticClusteringModel (#6337)
- Sped up Mutect2 reference confidence model with fast likelihoods model (#6457)
- Modified Fragment creation for Mutect2 to not fail for supplementary reads (#6327)
- Uniqify PG IDs in FilterAlignmentArtifacts (#6304)
- Fixed error in RealignmentEngine due to converting from exclusive to inclusive interval ends (#6404)
- Added an error message for no callable sites in Mutect2 (#6445)
- Changed filter reporting in Mutect2 (#6288)
- Fixed force-calling mode in M2 mito WDL (#6359)
- Pass the reference to the realignment filter in the Mutect2 WDL (#6360)
- Deleted the old orientation bias filter (#6408)
- Made callable sites a Long to avoid integer overflow (#6303)
GenomicsDB
- Move to GenomicsDB 1.2.0 (#6305)
  - Fixes an issue with GenomicsDBImport erroring out due to duplicate fields in the Info, Format, and/or Filter fields. (#6158)
  - Fixes an issue with GenomicsDBImport not completing for mixed ploidy samples (#6275)
  - This version uses a 64-bit htslib to workaround overflow issues when computed annotation sizes exceed the 32-bit integer space
Joint Calling
- GenotypeGVCFs: improved checking for upstream deletions in the GenotypingEngine (#6429)
  - Fixes rare cases where GenotypeGVCFs could emit a variant with a spanned allele (*), and a genotype that references the spanned allele, but fail to emit the upstream spanning variant.
- GenotypeGVCFs: Don't call the NON_REF allele in genotypes or ADs (#6437)
- Parse combined AS_QUALapprox values from older reblocked GVCFs properly (#6442)
- Added a force output sites argument to GenotypeGVCFs (#6263)
- Remove extraneous alleles in GenotypeGVCFs force-output mode (#6406)
CNV Calling
- Copy temporary files early in gcnvkernel to avoid inadvertent temporary directory cleanup. (#6297)
- Enabled streaming of counts.tsv/counts.tsv.gz files in gCNV CLIs. (#6266)
- Fixed shard index in PostprocessGermlineCNVCalls log message. (#6313)
- gCNV vcf cleanup (#6352)
- Index output VCFs for GCNV postprocessing (#6330)
Notable Enhancements
- VariantAnnotator is now out of beta (#6402)
- Concordance is out of beta (#6397)
- LeftAlignIndels now works for multiple indels (#6427)
- FilterVariantTranches can now handle cases where there are only SNPs or only indels, and not both (#6411)
- Added new read filters for NotProperlyPaired and for MateDistant (#6295)
- Made the .git directory optional during build (#6450)
Bug Fixes
- Handle zero-weight Gaussians correctly in VariantRecalibrator (#6425)
- Fixed the --invalidate-previous-filters argument in VariantFiltration to work as intended (ie., roll back all variants to unfiltered status) (#6412)
- Fixed a bug where SelectVariants takes forever on many-allelic somatic samples (#6446)
- Make sure SelectVariants outputs variants in correct order (assuming input vcf is correctly sorted) (#6444)
- Fixed a NPE crash in VariantEval when run with no intervals/reference (#6283)
- Fixed a NPE crash in FastaReferenceMaker (#6435)
- Fixed an out-of-bounds error in CountNs annotation (#6355)
- Fixed a bug in hardClipCigar function that caused incorrect cigar calculation (#6280)
- AnalyzeSaturationMutagenesis: fixed bug in codon calling for in-frame inserts (#6332)
Miscellaneous Changes
- Collect split read and paired end evidence files for GATK-SV pipeline (#6356)
- Add "PASS" filter line for ApplyVQSR and FilterMutectCalls (#6436)
- Added engine functionality for accessing the user defined intervals without merging them (#5887)
- Trim intervals loaded from interval files. (#6375)
- Propagate read group filters in ReadGroupBlackListReadFilter. (#6300)
- Modified ANDed read filter output message for readability (#6315)
- Clearly label the number of reads processed in the BaseRecalibrator log output (#6447)
- Clearly label the CountReads tool output (#6449)
- Improved the error messages for missing contigs in the reference (#6469)
- Avoid a copy and reverse operation in CigarUtils.isGood() (#6439)
- Fixed GenotypeAlleleCount's toString() method (#6376)
- Minor Funcotator WDL updates. (#6326)
- Added a getPairOrientation() method to GATKRead (#6420)
- Merged GATKProtectedVariantContextUtils methods into other classes (#6409)
- Deleted a lot of unused VCF constants (#6361)
- Deleted some unused genotyping code (#6354)
- Fixed incoherent unit test cases in allele subsetting utils (#6448)
- Add Python script executor error message for SIGKILL exit code 137. (#6414)
- Pip install pinned numpy. (#6413)
- Do not install R on travis, and only run the R tests on the Docker. (#6454)
- Fixes for IndexFeatureFile error reporting. (#6367)
- Temporarily remove dead Berkeley mirror to unblock builds. (#6422)
- Disable CNNVariantPipelineTest.testTrainingReadModel until failures are resolved. (#6331)
- Delete unused JsonSerializer (#6415)
- Delete empty file SparkToggleCommandLineProgram.java. (#6311)
Documentation
- Clarify the definition of the NON_REF allele (#6431)
- Clarify behavior of SplitIntervals for lists of adjacent intervals (#6423)
- Update docs to reflect the fact that TandemRepeat works with HaplotypeCaller (#5943)
- Update LeftAlignIndels documentation (#6177)
- Update hyperlink to new GATK forum page in the README (#6381)
- Add minValue/minRecommended value to ApplyBQSRArgumentCollection (#6438)
- Small README fixes (#6451)
- Fix some GATK doc issues (#6318)
- Update copyright date in LICENSE.TXT (#6383)
Dependencies
- Updated HTSJDK to 2.21.2 (#6462)
- Updated Picard to 2.21.9 (#6462)
- Updated Disq to 0.3.5 (#6323)
- Updated GenomicsDB to 1.2.0 (#6305)
- Updated TestNG to 7.0.0 (#5787)