The prediction of drug-target binding kinetics is a growing topic in computational biophysics of high relevance for biomedical research: about 15 years ago, it was realised that such kinetics often correlate better with drug efficacies than drug-target protein binding affinities^1-4. This insight poses a significant challenge for molecular dynamics (MD) simulation techniques, which can typically only access the lower microsecond range, as the unbinding of drugs from their protein receptors typically occurs on time scales extending to many hours. Attempts to bridge the time scale gap have led to the development of novel approaches in MD simulations^5-12. These approaches are aimed at, on the one hand, improving our understanding of the – often complex – mechanistic determinants of drug-target binding kinetics and, on the other hand, providing reliable and user-friendly tools to predict binding kinetic parameters that can be applied in the drug design pipeline^13-15. For example, biased or enhanced sampling MD simulation approaches enable the acceleration of binding and unbinding processes to investigate their mechanisms^16-22 and to provide a fast scoring of compounds according to their (un)binding characteristics^23-25. Further approaches that have been shown to hold promise are structural coarse-graining methods²⁶ and machine learning-based approaches^27-30. Compared to the much more established field of binding free energy calculations³¹, binding kinetics calculations face significantly larger challenges as regards adequate sampling, binding/unbinding path dependencies, force field accuracy for intermediate binding states, and the small available number of experimental datasets for computational benchmarks³². Following the insights of Copeland and others, comprehensive actions were undertaken to bundle theoretical as well as experimental^33-36 method development elucidating molecular kinetics. For example, the “Kinetics for Drug Discovery” IMI brought together researchers from both academia and industry to develop methods to measure and compute drug binding kinetic properties³⁷.

While the work on understanding (un)binding kinetics on a molecular level has continued and is present as sub-topics on computational conferences and workshops, the research of different groups has diverged, and no generally agreed-on “quality control” procedures exist for predictions of drug-target (un)binding kinetics. One critical cause is that no dedicated and repeated venue exists to bring together researchers of the field to discuss its state-of-the-art in depth, to identify current problems, to assess the problems that need to be solved in the future and to define “gold standards” for benchmark prediction methods. In this regard, it is indispensable to include experimentalists in these discussions for the generation of reliable experimental data sets on (un)binding characteristics of well-selected protein-ligand systems. Furthermore, connecting researchers from academia and industry is important to identify common goals and evaluate the applicability of currently existing approaches in drug discovery. Therefore, we propose a workshop that shall serve to gather the major players in the field of computational prediction of kinetics to focus the goals of the community, assess the current state-of-the-art and generate a generally agreed-on set of benchmark systems.