Thema der Dissertation:
Deep learning approaches for predicting pathogenic potentials of novel DNA and RNA sequences Thema der Disputation:
Discovering regulatory motif syntax in eukaryotic genomes with convolutional neural networks
Deep learning approaches for predicting pathogenic potentials of novel DNA and RNA sequences Thema der Disputation:
Discovering regulatory motif syntax in eukaryotic genomes with convolutional neural networks
Abstract: Binding of sequence-specific transcription factors (TFs) to short DNA motifs is one of the crucial mechanisms of gene regulation. While many motifs are known, the rules governing TF cooperativity are not fully understood. Avsec et al. (2021) train convolutional neural networks to predict binding profiles and interpret the learned patterns using post-hoc attribution methods. This generates robust motif representations, enabling identification of both previously unrecognized motifs and rules of their arrangement (syntax). In this talk, I will discuss how interpretable deep learning approaches can successfully model molecular phenotypes and support discovery of biologically relevant patterns in genomic data.
Zeit & Ort
30.06.2022 | 16:00