drugclaw/variant-analysis-tools

Variant Analysis Tools

Use this skill when the user provides a VCF or BCF and wants concrete counts, filtering, or mutation summaries instead of only database lookup.

Typical triggers:

• summarize the contents of a VCF or BCF
• count SNVs, indels, or structural variants
• filter by VAF, read depth, PASS status, or variant type
• exclude intronic or intergenic consequences from a local callset
• generate a machine-readable variant table before ClinVar, gnomAD, or dbSNP follow-up

Environment Check

which python3 || true
python3 - <<'PY'
mods = ["pysam"]
for name in mods:
    try:
        __import__(name)
        print(f"{name}: ok")
    except Exception as exc:
        print(f"{name}: missing ({exc})")
PY

Do not claim VCF analysis ran if pysam is unavailable.

Bundled Asset

• templates/variant_report.py

Preferred Workflow

1. Confirm which sample to read when the VCF is multi-sample.
2. Decide whether the user wants raw counts, filtered rows, or both.
3. Apply explicit filters for VAF, depth, PASS status, and consequence terms.
4. Export the filtered table plus a summary JSON.
5. If the user wants clinical significance or population frequency, hand the filtered rows to bio-db-tools for ClinVar, gnomAD, or dbSNP follow-up.

Quick Start

python3 templates/variant_report.py \
  --input cohort/sample.vcf.gz \
  --sample TUMOR \
  --pass-only \
  --min-vaf 0.05 \
  --min-depth 20 \
  --exclude-consequence intronic \
  --exclude-consequence intergenic \
  --output variants/sample_filtered.csv \
  --summary variants/sample_filtered.json

Structural-variant focused example:

python3 templates/variant_report.py \
  --input sv_calls.vcf.gz \
  --include-variant-type DEL \
  --include-variant-type DUP \
  --output variants/sv_subset.csv \
  --summary variants/sv_subset.json

Output Expectations

Good answers should mention:

• the exact variant file and sample used
• which filters were applied
• total records seen versus retained
• variant-type and consequence distributions
• top affected genes after filtering
• where the CSV and summary JSON were written

Related Skills

For ClinVar, Ensembl, gnomAD, or dbSNP lookups, activate bio-db-tools. For statistical testing or survival modeling on variant-derived burden tables, activate stat-modeling-tools or survival-analysis-tools. For target-level interpretation around genes hit by the variants, activate target-intelligence-tools.