varapp-backend-py 1.0.1.2

Motivation

Common filtering pipelines can be very tedious, including an annotation step, complicated forms, multiple exports to spreadsheets, merging and manual curation of theses spreadsheets. Although there already exist several tools to facilitate the annotation of lists of variants and apply basic filters such as frequency in a population, quality or location, Varapp has at least one of the following advantages over existing software:

• Security: It does not require to upload data to a cloud or external client; one can deploy the app on any local computer and keep the data private.

• Fast, non-trivial filtering based on family pedigree, including complex cases such as compound heterozygous, in less than a second for 500K variants.

• Reproducibility: It uses a relational database, adding the potential to cross information between experiments, reuse the results of previously studied data, and keep track of the annotation/versions used to produce the results.

• Reactive user interface: apply a filter and see the result immediately.

Use cases

Varapp is typically well-suited to cover the following two use cases, the result of which is obtained within a few seconds:

• Mendelian diseases

  We consider all variants in the exomes of one or several families affected by the same hereditary disease. We are searching for the variants that are susceptible to cause the disease.

  Varapp provides one-click filters for dominant, recessive, de novo, X-linked, and compound heterozygous cases.

• Gene panels

  We consider the same few genes (typically a few hundred) in a large series of independent individuals. We seek rare variants with strong impact, then inspect the individual(s) that carry them.

What Varapp is not

• … a variants caller:

  It starts after the variant calling has already been done, and works based on the VCF file that the latter produces.

• … adapted to full genome sequencing or million genomes projects:

  For the moment, the app is responsive with datasets up to 500’000 variants and a few hundred individuals, but performance issues are expected when dealing with millions of variants, as in the case of full genome sequencing, or with thousands of samples.