Variation Representation Specification (VRS) v2.0 is an approved GA4GH product

27 Mar 2025

A newly approved GA4GH product, the Variation Representation Specification (VRS) v2.0 establishes standardised descriptions of variant evidence across different sequence collections used by resources to advance a broader framework for sharing variant knowledge.

By Jaclyn Estrin, GA4GH Science Writer

The Global Alliance for Genomics and Health (GA4GH) is pleased to announce the product approval of the Variation Representation Specification (VRS) v2.0. In its second iteration, VRS v2.0 is a critical contribution to GA4GH’s suite of nearly forty products, all designed to enable responsible genomic and health data sharing to drive progress in research and advance human health outcomes.

Researchers who study genetic diseases, resulting from abnormalities in people’s DNA, look for the presence of genetic variation that may explain the diseases’ origins.

Genomics researchers across institutions use various conventions to represent and share knowledge about these variants. This means that researchers and molecular pathologists using data across these resources have the onerous task of piecing together these data across different contexts, leading to errors and missed information. 

In an effort to enable reliable and interoperable exchange of variant information among a diverse network of global variant evidence repositories, the Genomic Knowledge Standards (GKS) Work Stream developed the Variation Representation Specification (VRS; pronounced “verse”). VRS (v1.0 released in 2019) establishes a standardised computational language that enables integration of variant evidence across different sequence collections used by resources in order to advance a broader framework for sharing variant knowledge. 

Now the GA4GH Product Steering Committee has approved the product in its second iteration. Product development of VRS v2.0 was led by Alex Wagner (GKS Work Stream Co-Lead; Nationwide Children’s Hospital) and Larry Babb (Broad Institute of MIT and Harvard), with support from Work Stream Manager Beatrice Amos (Wellcome Sanger Institute) and Work Stream Co-Lead Melissa Cline (University of California, Santa Cruz).

Wagner said, “There was a need for us to form a standardised information model on variation that was semantically precise and computable, that could be used in a multitude of different contexts.”

VRS creates a standardised, interoperable way to represent genomic variant data across resources, so it can be integrated into one system. Relevant information can be quickly identified, accelerating the pace of variant interpretation. 

“We really needed a computational set of structures that allowed us to represent variation as it exists in a multitude of genomic evidence sources commonly used in variant interpretation and classification,” said Babb. “We wanted it to be computable, self-contained, and federated.”

Implementation of VRS v1.0 within several global genomic databases — including gnomAD (Genome Aggregation Database), ClinVar, CIViC (Clinical Interpretation of Variants in Cancer), and MaveDB — has garnered insights into the tool’s value and informed minor updates to the specification. 

The newly approved VRS v2.0 allows for additional descriptive metadata that commonly accompanies variants, making it more accessible for use across resources and with other GA4GH standards that currently use VRS v1.3, such as Phenopackets and Beacon. The product team is also working to align their standard with the newly approved refget Sequence Collections product to define variation with respect to a reference sequence. 

VRS v2.0 has made key advancements in using shared data structures to represent “small” variants (commonly described by other variant standards such as VCF [Variant Call Format] and HGVS [Human Genome Variation Society] nomenclature) alongside “structural” variants that include large-scale rearrangements of DNA. Importantly, VRS v2.0 also overcomes key limitations in the representation of variants in large regions of repeating sequences through a novel reference length encoding algorithm.

Babb said, “What we’re building is the structures. It’s like we’re giving people the LEGO blocks to build these different types of variation in a way that is computationally consistent and that will make sharing the data much easier and less prone to misrepresentation.” 

The team is now focusing on tools that make it easy for downstream resources to represent both sequence and structural variants, to ensure the pieces fit together in a cohesive data model to realise the vision of an interoperable resource ecosystem. A related product addressing Categorical Variation (Cat-VRS) is also in development.

Advancements in sequencing technologies and resources are galvanising the rapid generation of new sequences at a large scale. The VRS data model is vital to representing variation within these sequences, and in the view of the specification maintainers, is the core of a new generation of data representation formats.

“We are working towards a future where the conventions and models used by VRS underlie not just the VRS exchange protocol, but also the next generation of file formats and variant nomenclature,” said Wagner. “With the advancements in VRS v2.0, we are at a very exciting transition point where we may finally realise the goal of a common language for genomic variation.”

Implementation of VRS will strengthen the integrity of genomic data and advance a global understanding of how human genetic variation contributes to health and disease. Ultimately, this will drive progress in genomics research, as well as advancements in patient care and health outcomes.