RBP Footprint Grand Challenge: An evaluation of novel computational approaches to RNA-binding protein target prediction from structural data

The Froehlking team uses ADAR eCLIP data to train an generative model to predict ADAR binding site. Protein-bound nucleotides have been shown to display high eCLIP, and in some systems and for select nucleotides, low icSHAPE reactivities at the same time. However, obtaining both eCLIP and icSHAPE data is expensive. We thus built a machine learning model that can be used to generate sequences characteristic of protein-bound regions from experimental data, even if some information is missing, and to predict protein binding sites (details can be found in Materials and Methods of associated publication). ADAR eCLIP in K562 cells.