sVEP – accelerating variant annotation and curation

We’ve created a highly scalable cloud-native annotation framework for genomic variants that enables rapid turnaround for clinical tests.

The challenge

An individual’s genome can inform diagnostic and treatment choices. However, prioritising the actionable variants amongst the patient’s millions of genetic mutations is resource intensive.

Annotating a whole genome with clinical information and known functionalities is currently as time-consuming as reading the genome itself. This severely lengthens turnaround times for clinical tests.

Furthermore, with genomic information used in a wide range of application domains, relevant annotation sources can vary dramatically. This requires a highly flexible framework able to focus on the relevant information.

Our response

In collaboration with Pathology Queensland and QIMR Berghofer, we have created a fully cloud-native architecture that enables the rapid and flexible annotation of genomic variants.

Utilising AWS function compute, sVEP is a fully serverless framework that can be automatically deployed in a user’s cloud-account.

The framework can run over patient-individual VCF files or large cohort VCF files.

A graphic titled "Cloud-native Variant Annotation with sVEP. It shows steps in text boxes flowing left to right. 1st box VCF flows to 2nd box sVEP flows to Annotated csv which has three separate boxes flowing from it box a) clinic box b) pathology box c) research. The sVEP box has three vertical boxes flowing from it which say box a) sVEP plugin box b) sVEP plugin box c) sVEP plugin flowing below those boxes is on large text box which says Ensemble genes and Protein Effects on the right under a subheading of ensemble and on the the left Exac and gnomAD under a heading of THird Party

The framework has an orchestration engine that allows the selective execution of provided and external annotations plugins, making workflows highly customisable.

Benefits

Serverless VEP is estimated to be 98% faster than traditional VEP implementations as it can massively-parallelise the annotation task.

When not queried, the cloud set-up does not incur cost, making it sustainable for small pathology laboratories or research institutes.

An image showing serverless VEP is estimated to be 98% faster in processing time with a graph . sVEP can deliver 1480 extrapolations in 25 mins, whereas a standard VEP takes 24 hours to reveal 86400 extrapolations

Faster processing times are delivered by sVEP

You can view updates here