Overview of Features

Features are used to classify different groups of biomarkers with clinical relevance for cancer to support curation of structured data relevant to that type of biomarker. Features in CIViC are currently Genes, Fusions, and Factors, with other Features under development and planned for the future. Genes are specific regions (loci) in the genome, named using HUGO nomenclature (from Entrez Gene), and associated with structured data such as Ensembl transcripts, aliases, and protein domains. The original CIViC Variant data model was built entirely around Genes, and the majority of CIViC documentation is based on curation of variants directly associated to the Gene Feature class. But with that in mind, the data model for other Features broadly follows many aspects of the data model for Genes, including the relationship between Genes and Variants as seen below.

Feature Types in CIViC

Figure 1: Three Feature types exist in CIViC. Genes, which were the original Feature type, along with Fusions and Factors.

Generalizing the Gene-Variant data model to a Feature-Variant data model

Figure 2: The initial Gene-Variant data model of CIViC was generalized to Feature-Variant model, where each Feature has one or more Variants associated to it. New Variants are created by curators as new Evidence Items for that Variant are added to CIViC.

Features generalize a fundamental relation between a Gene and its Variants, so that for any instance of a Feature, multiple different variants may be associated with it. For example, a Fusion (Fusions are a class of Feature in CIViC) such as EML4::ALK can be associated with multiple Variants. In the case of Fusions, Variants may include different exon combinations, such as e6::e20 (EML4 exon 6 fused to ALK exon 20) and e20::e20, or the representative Fusion bucket variant when the specific breakpoints are not known (see section on Fusions and sections on other Feature types for more in-depth examples). With this generalization of the Gene-Variant data model to Features, it is especially useful to think of Variants in CIViC in terms of Feature-Variant pairs (e.g. BRAF V600E, HPV Positive, EGFR Mutation, EML4::ALK e6::e20, v::ALK Fusion). An overview of different Features in CIViC which are either implemented or under development is shown in the figure below:

Feature Types in the CIViC Data Model

Figure 3: The updated CIViC data model, which is partially implemented, contains four classes of Features. Genes were the first Feature implemented in CVIiC, and the data model was built around Genes and Gene based Variants, following a flexible model for the Variants. The Region feature type is currently under development, and consists of genomic regions. The Factor Feature type captures biomarkers with clinical relevance which are not tied to genes or regions such as kataegis, microsatellite instability, and the presence of HPV. This Feature type is coordinate free. The other Feature type is Fusion, encompassing transcript and regulatory fusions.