CITATION.cff

cff-version: 1.2.0
title: DeepRVAT
message: >-
  If you use this software, please cite it using the
  metadata from this file.
type: software
authors:
  - given-names: Brian
    family-names: Clarke
    orcid: 'https://orcid.org/0000-0002-6695-286X'
  - given-names: Eva
    family-names: Holtkamp
    orcid: 'https://orcid.org/0000-0002-2129-9908'
  - given-names: Hakime
    family-names: Öztürk
  - given-names: Marcel
    family-names: Mück
    orcid: 'https://orcid.org/0009-0000-3129-2630'
  - given-names: Magnus
    family-names: Wahlberg
    orcid: 'https://orcid.org/0009-0001-9140-2392'
  - given-names: Kayla
    family-names: Meyer
    orcid: 'https://orcid.org/0009-0003-5063-5266'
  - given-names: Felix
    family-names: Munzlinger
    orcid: 'https://orcid.org/0009-0005-1407-8145'
  - given-names: Felix
    family-names: Brechtmann
    orcid: 'https://orcid.org/0000-0002-0110-152X'
  - given-names: Florian Rupert
    family-names: Hölzlwimmer
    orcid: 'https://orcid.org/0000-0002-5522-2562'
  - given-names: Julien
    family-names: Gagneur
    orcid: 'https://orcid.org/0000-0002-8924-8365'
  - given-names: Oliver
    family-names: Stegle
    orcid: 'https://orcid.org/0000-0002-8818-7193'
identifiers:
  - type: doi
    value: 10.1101/2023.07.12.548506
repository-code: 'https://github.com/PMBio/deeprvat'
abstract: >-
  Integration of variant annotations using deep set networks
  boosts rare variant association genetics.
  Rare genetic variants can strongly predispose to disease,
  yet accounting for rare variants in genetic analyses is
  statistically challenging. While rich variant annotations
  hold the promise to enable well-powered rare variant
  association tests, methods integrating variant annotations
  in a data-driven manner are lacking. Here, we propose
  DeepRVAT, a set neural network-based approach to learn
  burden scores from rare variants, annotations and
  phenotypes. In contrast to existing methods, DeepRVAT
  yields a single, trait-agnostic, nonlinear gene impairment
  score, enabling both risk prediction and gene discovery in
  a unified framework. On 21 quantitative traits and
  whole-exome-sequencing data from UK Biobank, DeepRVAT
  offers substantial increases in gene discoveries and
  improved replication rates in held-out data. Moreover, we
  demonstrate that the integrative DeepRVAT gene impairment
  score greatly improves detection of individuals at high
  genetic risk. We show that pre-trained DeepRVAT scores
  generalize across traits, opening up the possibility to
  conduct highly computationally efficient rare variant
  tests.
license: MIT