vcfcache 0.3.3

Cache-based VCF annotation accelerator

VCFcache – Cache once, annotate fast

Cache common variants once, reuse them for every sample. VCFcache builds a normalized blueprint, annotates it once, and reuses those results so only rare/novel variants are annotated at runtime.

---

What’s new

• Single lean Docker image: ghcr.io/julius-muller/vcfcache:<version>
• Caches/blueprints live on Zenodo and are resolved by alias via public_caches.yaml
• Alias schema:
  • Blueprint: bp-<genome>-<source>-<release>-<filt>
  • Cache: cache-<genome>-<source>-<release>-<filt>-<tool>-<tool_version>-<preset>
• Manifest entries carry version (vcfcache version used), no image tags
• Bundled bcftools 1.22; no Nextflow/JVM required

---

Quick start (pip install)

Requirements: Python >= 3.11, bcftools >= 1.20

# Install vcfcache
pip install vcfcache

# Run comprehensive demo (tests all 4 commands)
vcfcache demo

# Use it
vcfcache --help

Note: bcftools is a runtime dependency and must be installed separately:

• Ubuntu/Debian: sudo apt-get install bcftools
• macOS: brew install bcftools
• Conda: conda install -c bioconda bcftools

The demo command runs a complete workflow testing:

1. blueprint-init - Create cache from variants
2. blueprint-extend - Add more variants
3. cache-build - Annotate the cache
4. annotate - Use cache to annotate sample VCF

---

Quick start (from source)

Requirements: Python 3.13+, uv, bcftools >= 1.20

git clone https://github.com/julius-muller/vcfcache.git
cd vcfcache
uv venv .venv && source .venv/bin/activate
uv pip install -e ".[dev]"
vcfcache --help
python -m pytest tests -q   # optional

---

Quick start (Docker + public cache)

The CLI can fetch caches from Zenodo by alias using the bundled public_caches.yaml.

docker pull ghcr.io/julius-muller/vcfcache:v0.3.0

docker run --rm -v $(pwd):/work ghcr.io/julius-muller/vcfcache:v0.3.0 \
  annotate \
    -a cache-hg38-gnomad-4.1joint-AF0100-vep-115.2-basic \
    --vcf /work/tests/data/nodata/sample4.bcf \
    --output /work/out \
    --force

List public caches:

vcfcache list --public-caches

Show the recorded annotation command without running it:

vcfcache annotate -a cache-hg38-gnomad-4.1joint-AF0100-vep-115.2-basic --show-command

---

Build your own cache

1. Normalize/deduplicate into a blueprint:

vcfcache blueprint-init \
  --vcf gnomad_subset.bcf \
  --output ./cache \
  -y params.yaml \
  --force

2. Annotate the blueprint (cache-build) with your recipe:

vcfcache cache-build \
  --name vep_custom \
  --db ./cache \
  -a annotation.yaml \
  -y params.yaml \
  --force

3. Use the cache on samples:

vcfcache annotate \
  -a ./cache/cache/vep_custom \
  --vcf sample.bcf \
  --output ./results \
  --force

---

Manifest & aliases

• Manifest: public_caches.yaml
• Fields (minimum): alias, type (blueprint|cache), version, genome, source, release, filt, doi, updated_at, md5, plus tool/tool_version/preset for caches.
• Aliases:
  • Blueprint: bp-hg38-gnomad-4.1joint-AF0100
  • Cache: cache-hg38-gnomad-4.1joint-AF0100-vep-115.2-basic
• vcfcache annotate -a <alias> auto-downloads from Zenodo and extracts to ~/.cache/vcfcache/caches/<alias>.
• vcfcache pull --doi <doi> --dest <dir> for manual download/extract.
• vcfcache push --cache-dir <dir> uploads to Zenodo (requires ZENODO_TOKEN; uses real Zenodo by default).

---

Configuration (YAML)

Two files, stored with the cache for reproducibility:

1. params.yaml — resource paths/tools:

bcftools_cmd: "bcftools"         # bundled bcftools by default
annotation_tool_cmd: "vep"
tool_version_command: "vep --version"
temp_dir: "/tmp"
optional_checks: {}

2. annotation.yaml — immutable annotation recipe:

annotation_cmd: |
  ${params.bcftools_cmd} view ${INPUT_BCF} | \
  ${params.annotation_tool_cmd} --offline --cache \
    --dir_cache ${params.vep_cache} --format vcf --vcf \
    -i STDIN -o STDOUT | \
  ${params.bcftools_cmd} view -o ${OUTPUT_BCF} -Ob -W
must_contain_info_tag: CSQ
required_tool_version: "115.2"
optional_checks:
  vep_cache_version: "115"
  genome_build: "GRCh38"

Special variables: ${INPUT_BCF}, ${OUTPUT_BCF}, ${AUXILIARY_DIR}, and ${params.*}.

---

Testing

• From source: python -m pytest tests -q
• In Docker (dev stage):
  docker build --target test -f docker/Dockerfile.vcfcache -t vcfcache:test .
docker run --rm vcfcache:test /opt/venv/bin/python -m pytest tests -q

---

Docker

• Lean runtime: ghcr.io/julius-muller/vcfcache:v0.3.0 (bundled bcftools 1.22)
• Multi-stage build: docker/Dockerfile.vcfcache
  • --target test installs dev deps and can run pytest
  • --target final is the shipped image (small)

---

Notes

• No Nextflow/JVM needed.
• Caches are portable tarballs; manifest version is informational to track which vcfcache built them.
• bcftools is bundled and used by default; override via bcftools_cmd if you need a different binary.