torchcell.verification.verify_fitness_dataset

verify_fitness_dataset(records: Sequence[dict[str, Any]], *, dataset_name: str, provenance: Provenance, expected_count: int, resolve_gene_name: Callable[[str], GeneNameResolution] | None = None, sgd_genes: set[str] | None = None, gene_universe_label: str = 'reference', min_containment: float = 0.9) → VerificationReport

Run the L0-L4 record-level gate for a single-mutant fitness dataset.

sgd_genes turns on the L4 gene rules (aggregate containment + per-record genome membership), gene_universe_label is what the containment row calls that universe (the host’s, not always S288C’s), and resolve_gene_name the annotation half of the canonical-name rule; without them the caller owns L4 (fitness_gene_set() is the overlap key).