Features

Variant Object Types

AnyVar enables the registration and retrieval of a number of different variation object types, beginning with core elements of the GA4GH Variation Representation Specification (VRS):

• VRS Allele

• VRS SequenceLocation

• VRS SequenceReference

Variant Translation

AnyVar implements a Translator abstraction that can be used to ingest free-text variation expressions of known nomenclatures. By way of the VRS-Python translator module, the following kinds of expressions are supported:

Expression Type	Example
HGVS	NC_000007.14:g.140753336A>T
SPDI	NC_000007.14:140753335:A:T
gnomAD/VCF	7-140753335-A-T

VCF Ingestion and Annotation

AnyVar can ingest and register all variants (and reference alleles) contained within a Variant Call Format (VCF) file, and return a file copy with variation IDs included as INFO field properties.

Object Extensions

Registered variation objects can be associated with extensions, which provide a flexible mechanism for attaching additional, lightweight metadata to a variation. An extension consists of: a name (str), describing the type of metadata, and a value, which may be any JSON-serializable object.

Extensions are intended to mirror the concept of extensions in VRS: they capture auxiliary information that is closely associated with the variation itself, but not part of its core, identity-defining representation. Extensions are deliberately unstructured and permissive in order to support a wide range of use cases without requiring schema changes.

Examples of appropriate uses include:

• registration metadata (e.g., timestamps, provenance identifiers)

• cross-references to external resources

• links to associated samples, patients, or datasets

• simple flags or attributes relevant to the stored object

Note

Extensions are not intended for storing large, complex, or highly interpretive data. In particular, they should not be used for:

• detailed evidence or annotation records

• clinical interpretations or assertions

• large structured payloads or documents

Such information is better represented in dedicated data models or external systems.

Variant Mappings

Mappings can be used to register specific modes of relationship between variations, such as reference assembly liftover. The VariationMappingType enum provides the supported kinds of relationships:

class anyvar.core.metadata.VariationMappingType(value)

Supported mapping types between VRS Variations.

Warning

Currently, use of mappings outside of the liftover relation are experimental, and these parameters are subject to change. The following describes current intentions, and may or may not be validated within AnyVar.

• LIFTOVER_TO: Genomic-to-genomic coordinate transformation. Use when mapping a variation between two reference sequences of the same molecule type, typically genomic DNA.

• TRANSCRIBE_TO: Genomic-to-transcript (RNA/cDNA) projection. Use when projecting a genomic DNA object onto a transcript sequence (RNA or cDNA). This accounts for splicing, transcript strand orientation, and transcript-specific exon structure.

• TRANSLATE_TO: Transcript-to-protein projection. This entails codon interpretation, amino acid substitution/insertion/deletion/extension, and protein coordinate changes.

LIFTOVER_TO = 'liftover_to'

TRANSCRIBE_TO = 'transcribe_to'

TRANSLATE_TO = 'translate_to'

Stateless Mode

Optionally, an AnyVar server can be configured as stateless, to provide its variant translation and bulk file annotation functions without any persistent object registration backend. This utilizes the NoObjectStore class in place of a relational data backend.