broadinstitute/gatk 4.0.4.0

on GitHub

Highlights of this release include major performance improvements to MarkDuplicatesSpark, better sensitivity and precision in STR (short tandem repeat) contexts for Mutect2, support for a "genotype given alleles" mode in Mutect2, dbSNP support for Funcotator, and several important bug fixes to CombineGVCFs.

As usual, a docker image for this release can be downloaded from https://hub.docker.com/r/broadinstitute/gatk/

Full list of changes in this release:

• MarkDuplicatesSpark

  • New, optimized version of the tool with greatly improved performance and scalability (#4656)
  • Note that this tool is still marked as beta, and has a number of known issues. The current version is suitable for evaluation/profiling purposes only.

• Mutect2 improvements

  • Added a GGA (genotype given alleles) mode activated via the --genotyping-mode GENOTYPE_GIVEN_ALLELES and --alleles arguments (#4601)
  • Better sensitivity and precision in STR (short-tandem repeat) contexts (#4690)
  • New, supported Mutect2 NIO-enabled WDL that works in Firecloud (#4710)
  • Better default AF for M2 tumor-normal mode (#4690)
  • Restored explicit PASS (as opposed to empty) filter in Mutect2 (#4644)
  • Fixed Mutect2 failure for germline resource without AF (#4607)
  • Fixed a bug in the Mutect2 WDL bamout where scatters with overlapping assembly regions failed (#4613)
  • Fixed extra filtering args being deactivated in Mutect2 WDL due to typo

• CombineGVCFs: several important bug fixes

  • ReferenceConfidenceVariantContextMerger fixes for spanning deletions, and use the correct types for the median calculation. (#4680)
  • Handle trailing reference blocks correctly (#4615)
  • Fix and test for calculating intermediate band interval start locations. (#4681)

• Funcotator

  • Added dbSNP support via a new VcfFuncotationFactory. (#4593)
  • Fixed the refContext annotation. (#4605)
  • Fixed calculation of GC content to be correct. (#4608)
  • Fixes for HG38 exception and better logging. (#4563)
  • Note: only datasource releases 1.2.20180329 and later will work with this version of Funcotator

• HaplotypeCaller: Fixed a bug that caused the --comp and --input-prior arguments to not be settable by the user (#4703)

• CNNScoreVariants: Better numerical consistency between python and java, and transpose bug fix (#4652)

• CNV Tools

  • A new framework to support automated evaluation of GATK CNV (#4276)
  • Enabled zero eigensamples to be specified for CreateReadCountPanelOfNormals (#4502)
  • Exposed maximum chunk size in CNV panel of normals. (#4528)
  • Changed CNV PoN to filter on equality to interval median percentile. (#4503)

• SV Tools

  • Breakpoint location and type inference unit (#4562)
  • Scaffold local assemblies (#4589)
  • Use the latest version of fermilite jni (#4622)
  • Update sv scripts to only copy a single bam file and index, and respect project parameter (#4646)
  • Various bug fixes (#4670) (#4623)

• Added GCS (Google Cloud Storage) output support to the following tools: ApplyBQSR, SplitNCigarReads, ClipReads, LeftAlignIndels, RevertBaseQualityScores, and UnmarkDuplicates (#4695) (#4424)

• Mark the --disable-tool-default-read-filters argument as advanced, and add a warning to its documentation string (#4671)

  • Many tools do not function correctly without their default read filters turned on, so this argument is intended only for advanced users who know what they're doing!

• ParallelCopyGCSDirectoryIntoHDFSSpark: allow the tool to take a filename glob to subset files to copy (#4624)

• Picard: updated to version 2.18.2 (#4676)