CAGS: Deep neural networks and DNA variant effect predictions in humans

Speaker:  Ágúst Egilsson, NATERA Inc.

Title:  Deep neural networks and DNA variant effect predictions in humans

Abstract: 

In this applied session we will discuss emerging approaches that rely on the generalizability and explainability of deep neural networks for understanding genomic variants and DNA regulatory activity in the non-coding regions of the human genome. Currently, these approaches rely on the use of convolutions or transformers and variational autoencoders and are achieving state-of-the-art performance comparable to experimental results in some use cases. During the talk we will detail the network structures and explore the recent history of the approaches. Keywords: Deep neural networks, convolutions, transformers, multi head attention, autoencoders, DNA, genome, variant classification.

PDF of Abstract

Time: Friday, December 3, 2021, 12:30pm – 1:30pm