Blind protein-ligand docking with diffusion-based deep generative models

Deep learning techniques have led to significant advancements in data-driven modeling of biomolecular structure, function, and interactions. We develop a diffusion-based deep generative model for blind protein-ligand docking that learns a probability distribution over ligand poses conditioned on the target protein structure. As the space of ligand poses are described by a non-Euclidean manifold, we map this manifold to the product space of the degrees of freedom (translational, rotational, and torsional) involved in docking and develop an efficient diffusion process on this space.