PILOT: Equivariant Diffusion for Pocket Conditioned De Novo Ligand Generation with Multi-Objective Guidance Via Importance Sampling

The generation of ligands that both are tailored to a given protein pocketand exhibit a range of desired chemical properties is a major challenge instructure-based drug design. Here, we propose an in-silico approach for thede novo generation of 3D ligand structures using the equivariantdiffusion model PILOT, combining pocket conditioning with a large-scalepre-training and property guidance. Its multi-objective trajectory-basedimportance sampling strategy is designed to direct the model towards moleculesthat not only exhibit desired characteristics such as increased bindingaffinity for a given protein pocket but also maintains high syntheticaccessibility. This ensures the practicality of sampled molecules, thusmaximizing their potential for the drug discovery pipeline. PILOT significantlyoutperforms existing methods across various metrics on the common benchmarkdataset CrossDocked2020. Moreover, we employ PILOT to generate novel ligandsfor unseen protein pockets from the Kinodata-3D dataset, which encompasses asubstantial portion of the human kinome. The generated structures exhibitpredicted IC_50 values indicative of potent biological activity, whichhighlights the potential of PILOT as a powerful tool for structure-based drugdesign.