VCF Normalization Profile (vcf_normalization_v1)

vcf_normalization_v1 is a Tier 1 semantic profile that verifies VCF normalization steps: variants must remain semantically identical while being left-aligned and minimally represented.

What it checks

• Variant preservation: canonicalized original equals canonicalized normalized (no gain/loss of truth).
• Left-alignment/minimality: no redundant shared prefix/suffix remains; locus updates reflect any prefix trimming.
• Ref/alt consistency: ref/alt remain valid after trimming.
• Idempotence: re-normalizing an already normalized variant is a fixed point.

Schema shape (summary)

schemas/variant/vcf_normalization_v1.schema.json:

• input_vcf_hash, normalized_vcf_hash: content hashes (e.g., sha256:<hex>).
• reference_fasta_hash (optional).
• variants[] entries:
• locus: chrom, pos
• original: ref, alt[]
• normalized: pos, ref, alt[]
• operations: optional list like left_trim, right_trim, left_align, split, merge, pad_left, pad_right.
• summary: optional counts (trimmed/split/invalid).

CLI usage

veribiota check vcf vcf_normalization_v1 path/to/input.json --snapshot-out sig.json --compact

Output: JSON verdict with variant_count, normalized_ok_count, all_variants_normalized, hashes, theorem IDs (VB_VCF_001, VB_VCF_002), and engine metadata.

Snapshot: --snapshot-out writes a snapshot_signature_v1 document binding the input hash, schema hash, theorem IDs, and build metadata to the result.

Example input

{
  "input_vcf_hash": "sha256:aaaaaaaa...aaaaaaaa",
  "normalized_vcf_hash": "sha256:bbbbbbbb...bbbbbbbb",
  "variants": [
    {
      "locus": { "chrom": "chr1", "pos": 1 },
      "original": { "ref": "AC", "alt": ["A"] },
      "normalized": { "pos": 2, "ref": "C", "alt": [""] },
      "operations": ["left_trim"]
    }
  ]
}

Run:

veribiota check vcf vcf_normalization_v1 examples/vcf_norm.json --snapshot-out sig.json

• Verdict: all_variants_normalized: true, status: "passed".
• Signature: sig.json in snapshot_signature_v1 format for provenance.