While binding affinity is typically the primary focus in the early stages of drug discovery, protein-ligand kinetics can strongly influence efficacy, dosing, toxicity, and selectivity. But although computational methods for binding affinity are abundant, tools for kinetics have so far been limited, with most approaches relying on computationally intensive molecular dynamics simulations. In contrast, Koffee™ Unbinding Kinetics delivers physics-based residence time estimates in just 1—2 minutes per complex on a GPU, without any machine learning or target-specific fine-tuning.

Our approach ranks chemically diverse ligands — including those with different scaffolds and charge states — and spanning in size from fragments to large macrocycles. In contrast, most binding affinity methods only perform reliably on congeneric series.