Large-scale docking and generative modeling in CACHE challenge #1

Abstract:

Identifying binders for new protein targets or domains is a large challenge for drug discovery projects and a large fraction of the proteome has not been liganded to date. Virtual screening can provide fast access to chemical matter, however, it is often challenging to properly validate different approaches. The community-wide blind challenge CACHE provides a unique opportunity to benchmark different methods without biases in setups that are close to challenges in real-life drug discovery projects. Here, we describe a combination of physics-based docking and generative modeling as an approach to screen large virtual spaces such as Enamine REAL using and how we applied this in CACHE challenge #1 to identify a novel binder targeting the WDR domain on LRRK2. We will also discuss a retrospective evaluation of an active learning docking approach.