v0.4.0

The most significant change is that datasets are now always stored as memory maps, without loss of speed. Training that previously required 64 Gb of RAM now requires 8 Gb.

What's Changed

• Deepsomatic pipeline for comparisons by @davidbenjamin in #231
• Call-caching hack in WDL scripts because Terra has been buggy by @davidbenjamin in #232
• Edited paper manuscript some more by @davidbenjamin in #234
• More extensive use of memory maps; huge reduction in memory requirements by @davidbenjamin in #235

Permutect v0.4.0 is designed to be used with the following public resources/versions:

• gnomad germline resource: gs://broad-dsp-david-benjamin/af-only-gnomad-4.1.0-hg38.vcf.gz
• GATK docker: us.gcr.io/broad-gatk/gatk:4.6.0.0
• GATK override: gs://broad-dsp-david-benjamin/gatk-builds/gatk-4-5-2025.jar
• Mutect2 WDL: https://app.terra.bio/#workflows/davidben/mutect2/20/wdl (this is used as a subworkflow in the training dataset and variant calling workflows in this repo)
• Permutect docker image: us.gcr.io/broad-dsde-methods/davidben/permutect:v0.4.0
• Permutect artifact model: gs://broad-dsp-david-benjamin/permutect/permutect-v0.4.0-artifact-model.pt

The GATK override and the Mutect2 WDL subworkflow are not yet in a GATK release but are merged into the gatk master branch as of 4/5/2025.

The docker image was generated from the current Dockerfile in this repo with the command:
docker build -t us.gcr.io/broad-dsde-methods/davidben/permutect:v0.4.0 --platform linux/amd64,linux/arm64 .

Full Changelog: v0.3.0...v0.4.0