What AI says about KarmaDock

KarmaDock is a deep learning framework for ultra-large library ligand docking that enables fast, accurate structure-based virtual screening in drug discovery. It uses a three-stage model that encodes protein and ligand representations, updates ligand poses with E(n) equivariant graph neural networks and self-attention, and scores binding strength with a mixture density network. The approach has been validated on benchmark datasets and in a real-world virtual screening that identified experimentally validated inhibitors of leukocyte tyrosine kinase (LTK).