Atlas Variant Database Methodology
Atlas is the patient-independent variant classification database built and maintained by Helena Bioinformatics. It stores one governed record for each eligible normalized variant so the same patient-independent annotation and classification work can be reused consistently.
Atlas is not a patient registry, a VCF archive, or a copy of an upstream assertion database. This page documents its scientific scope, identity rules, release controls, and interpretation boundaries without publishing proprietary implementation details.
Published and reviewed by Helena Bioinformatics · Last reviewed 17 August 2026
Where Atlas Is Used
Inside the Folklore app
Folklore checks Atlas read-only for a current exact match. A valid match supplies reusable patient-independent annotation and classification evidence, then rejoins the normal workflow before phenotype, genotype, inheritance, family, and other case context are evaluated.
Public variant interpretation
Anyone can search a supported GRCh38 nuclear variantthrough Folklore's public interface. An exact current Atlas match can supply the patient-independent result; a miss or failed validation continues through the standard interpretation path rather than returning stale evidence.
How Atlas Is Built
Atlas treats normalized variant identity as the unit of the database. Source repetition does not create multiple records, and a published generation cannot mix classifications produced under different scientific releases.
Define the eligible variant scope
Atlas admits normalized GRCh38 nuclear single-nucleotide variants and small insertions or deletions. Mitochondrial variants and structural variants remain in their dedicated interpretation frameworks.
Establish one canonical identity
Each variant is represented by its chromosome, position, reference allele, and alternate allele after normalization. Equivalent source representations converge on the same identity.
Assemble governed evidence
The variant is annotated with current transcript and consequence information and the available population, clinical, prediction, conservation, dosage, phenotype, and gene-disease reference evidence.
Classify without patient context
The Helena classifier applies the governed ACMG/AMP evidence framework without patient phenotype, genotype, inheritance, segregation, family, or case-specific evidence.
Bind every row to one release
One Atlas generation uses one coherent classification release that binds the rules, reference evidence, normalization policy, annotation policy, and record contract used to produce its rows.
Validate and publish
A complete candidate is checked for duplicate identities, missing required data, unsupported classes, mixed releases, and patient-derived content before it can replace the active generation.
What an Atlas Record Represents
Every accepted record is a patient-independent evidence capsule for one normalized variant. The exact fields available depend on the applicable evidence sources, but the release contract keeps their interpretation consistent across the generation.
Canonical identity
GRCh38 coordinates and alleles, together with supported public variant notations and identifiers when available.
Annotation context
Gene, transcript, consequence, predicted impact, and other patient-independent attributes needed for evidence review.
Reference evidence
Governed population, clinical assertion, prediction, conservation, dosage, phenotype, and gene-disease signals with release provenance.
Classification capsule
One of the five ACMG classes, the applied evidence criteria, and the classification-release provenance needed to interpret that result.
Classification and Release Governance
Atlas uses the five ACMG sequence-variant classes: Pathogenic, Likely pathogenic, Uncertain significance, Likely benign, and Benign. The framework follows the ACMG/AMP sequence variant interpretation guidelines together with applicable ClinGen classification guidanceand Helena's documented safeguards.
When classification behavior, governed references, normalization, or the record contract changes materially, Atlas is rebuilt as a separate candidate. The candidate must pass completeness, identity, class, release-consistency, and privacy gates before it becomes active. The existing generation remains available until the replacement has passed those checks.
Patient Data Boundary
Atlas does not contain patient identifiers, cases, sessions, genome fingerprints, genotypes, original VCFs, family relationships, or patient phenotype. Patient-specific evidence is evaluated by the owning Folklore workflow after Atlas lookup and is not written back into the Atlas record.
Coverage and Limitations
- Atlas covers eligible GRCh38 nuclear SNVs and small indels present in the governed Atlas variant universe; it is not a catalog of every theoretically possible variant.
- Absence from Atlas is not evidence that a variant is benign, irrelevant, or unclassifiable. Folklore continues through the standard annotation and classification path.
- Mitochondrial and structural variants use their dedicated frameworks and are not classified by Atlas.
- A stored class is patient-independent and release-specific. It can change when evidence, guidance, or classification behavior changes.
- Atlas supports evidence review; it does not diagnose a patient or determine reporting, treatment, surveillance, penetrance, or clinical actionability.
Current total and class-level counts are maintained on the Atlas product page rather than duplicated here.
Access and Related Resources