On Friday, May 19th at 2pm Joschka and Michael will present their work on using docking data for structural models in the BICS (bioinformatics council Saarbrücken) talk series.

Abstract

Modern drug discovery pipelines increasingly rely on machine learning as a central component. They enable the exploration and evaluation of large chemical spaces. Conceptually, structural models are particularly interesting. Their application, however, is often hindered by the availability of target-ligand complex structures. We study the generation of complex data sets using available assay data to tackle this issue. We focus on kinases, because their binding mechanisms are well-understood and hence are well-suited to docking. Furthermore, they represent a highly relevant pharmaceutical target. Our evaluation shows that binding affinities can be predicted with higher precision by models that take synthetic binding poses into account.