Publications in peer-reviewed journals

DKFZ, EMBL, TU Muenchen Develop AI Model DeepRVAT for Rare Variant Testing

Sep 26, 2024, 06:12 AM

Researchers at DKFZ, EMBL, and TU Muenchen have developed a new AI model called DeepRVAT. This model leverages deep set neural networks to integrate variant annotations, enhancing the accuracy of rare variant association testing. DeepRVAT allows for more precise distinction of individuals at high risk of disease and helps identify genes involved in disease development. The model first accounts for nonlinear effects from rare variants on gene function and then models variation in traits as linear functions of estimated gene impairment scores. This breakthrough, developed in collaboration with Holtkamp_Eva and brianfclarke, pushes the boundaries of personalized medicine by providing more accurate predictions of the effects of rare genetic variants. The model also harnesses variant annotations from models like AlphaMissense.