65. Creating a common language for categorical variants

Abstract

Categorical variants serve an indispensable role in genomic knowledgebases to represent discrete classes of genomic variation. Despite their role linking classes to genomic evidence, their usefulness is undermined by confusion about and inconsistent use of categorical variant labels and class membership criteria.

Multiple labels may exist for the same class of variants, such as a {gene} deletion interchangeably labeled as a {gene} loss. In other cases, a single label is applied to multiple distinct classes of variants, as when {gene} loss could denote either a copy number loss or a loss-of-function. These cases underscore the challenge of our current state where categorical variants are fraught with ambiguity.

The Global Alliance for Genomics and Health (GA4GH) Categorical Variation Working Group is developing the Categorical Variation Representation Specification (Cat-VRS) to alleviate these issues. The Cat-VRS describes categorical variants by the concrete properties common to the class of assayed variants they represent. Applying the Cat-VRS to the example above, a sequence variant showing a lack of the gene indicates deletion of that gene and is therefore a copy number loss. This is distinct from a loss of gene product with no corresponding sequence deletion that unambiguously characterizes a loss-of-function variant.

As an unambiguous and computable representation standard, Cat-VRS will improve knowledgebase search and curation, and support the development of automated tools for knowledgebase harmonization and clinical variant analysis. This specification aims to supersede VRSATILE as a representation standard for categorical variation.