Integration of variant annotations using deep set networks boosts rare variant association testing.

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Rare genetic variants significantly impact phenotypes but pose statistical challenges in genetic analyses due to low carrier numbers and multiple testing burdens.
• Existing methods lack data-driven integration of rich variant annotations for powerful rare variant association tests.

Purpose of the Study:

• To introduce Deep Rare Variant Association Testing (DeepRVAT), a novel model for analyzing rare genetic variants.
• To enable well-powered rare variant association tests by integrating variant annotations in a data-driven manner.
• To facilitate both gene discovery and trait prediction using rare variant data.

Main Methods:

• Developed DeepRVAT, a set neural network model.
• The model learns a trait-agnostic gene impairment score from rare variant annotations and phenotypes.
• Applied DeepRVAT to whole-exome sequencing data from the UK Biobank for 34 quantitative and 63 binary traits.

Main Results:

• DeepRVAT demonstrated substantial improvements in gene discovery across various traits.
• The method enhanced the detection of individuals at high genetic risk.
• Showcased calibrated and computationally efficient rare variant tests at biobank scale.

Conclusions:

• DeepRVAT offers a powerful approach for rare variant genetic analyses.
• The model aids in discovering genetic risk factors for human diseases.
• DeepRVAT advances the integration of variant annotations for genetic studies.