Speaker: Elias Boudjella.
Abstract: This presentation describes a deep learning approach for predicting HIF binding sites from DNA sequences. It begins with the biological background and the construction of the datasets using ATAC-seq and ChIP-seq data from different cell types. It then presents the training of DNABERT-2 models and the performance of a combined model. Finally, mutation-based analyses are used to investigate the sequence features driving the predictions and to identify the nucleotides that contribute most to HIF binding prediction.