Intracellular crowding plays a crucial role in the behaviour of proteins,
influencing their motion, kinetic and equilibrium properties. In this work, we
present an adaptive model to simulate the association process between two
biomolecules in a crowded environment by Brownian dynamics, and thereby
compute the bimolecular association rate constant. In this model, interactions
are represented with atomistic resolution when crowder molecules are
nearby and with a coarser force field when they are farther away. We applied
this model to two contrasting protein-ligand systems with varying crowder
types and densities. Markov State modelling of the encounter trajectories
was used to analyse the binding pathways. The results reveal competing
effects resulting in a non-monotonic dependence of the association rate on
crowder concentration. While the presence of the crowders can hinder binding
by reducing the translational diffusion of the molecules in the medium
and obstruction of binding sites, the crowders can also enhance association
via caging, channeling, and other mechanisms. We are currently working on
incorporating physics-informed machine learning to improve the calculation
of short-range intermolecular forces, to increase the accuracy of our simulations.