Answers are grounded in this paper and cite their sources.
In the drug discovery field, drug affinity has long been considered the gold standard for designing small compounds with therapeutic activity. However, since the beginning of the 21st century, kinetic parameters, which are directly related to compound efficacy, have increasingly been recognized as central to the rational design of new drugs. However, such data are scarce, while the experimental determination of the association and dissociation rate constants, k on and k off , is costly and time-consuming. Molecular simulations, such as Molecular Dynamics (MD) calculations have been proposed to predict those kinetic constants, despite known limitations. With the rapid development of AI-based tools, researchers are now proposing advanced methodologies that leverage both data-driven learning from existing datasets and physics-based modelling through MD simulations. This review presents a critical overview of MD-based approaches for estimating protein-ligand binding free energies and rate constants. It then examines data-driven prediction methods and ML-assisted molecular simulations, with particular attention to validation and current limitations. Finally, it discusses the crucial need for new data with respect to the FAIR principles and methodological developments needed to further improve prediction quality.
No takes yet. Share an insight, caveat, or question.
Etchebest et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: