183P - SomaVar: A web application for somatic variant storage, annotation and classification

Background

Personalized targeted cancer treatment strongly depends on genetic tests in tumor tissue. Meanwhile, it is essential to understand the oncogenicity of a large number of somatic and germline variants to identify potential drug targets, which can be used for further treatment. An adopted classification scheme to guide variant oncogenicity assessment has to be easy, fast and reproducible. In this study we implemented a full-fledged database SomaVar and web-based frontend for somatic variant upload, annotation and classification using the VICC classification guidelines.

Methods

The frontend provides a multitude of utility functions which include an advanced variant search, user defined variant lists, automatic classification criteria selection and extensive administrative utilities. SomaVar supports variant annotation including in-silico pathogenicity prediction scores, splicing prediction scores, third party classifications, population scores and protein domains. SomaVar has a semi-automatic classification algorithm for oncogenicity based on the VICC standards and allows harmonized consensus classification between different users.

Results

Using a list of variants already classified by experts we show that the automatic classification algorithm provides a high-quality preselection for most VICC criteria. This not only increases the speed at which experts can classify variants, but also increases quality by mitigating human error. We also compared our knowledgebase to OnkoKB. A large overlap was found within the both groups of variants of likely oncogenic and of unclear oncogenicity variants. Interestingly, divergent classification was observed for variants located within splice regions and for benign variants listed in gnomAD at high allele frequencies.

Conclusions

To date, our knowledgebase harbors over 20.000 classified somatic variants and continues to grow. This open-source knowledgebase is publicly available online.

Editorial acknowledgement

Clinical trial identification

Legal entity responsible for the study

The authors.

Funding

Has not received any funding.

Disclosure

All authors have declared no conflicts of interest.