Abstract
The long timescale associated with ligand residence times renders their computation challenging. Therefore, the influence of factors like solvation and steric hindrance on residence times are not fully understood. Here, we demonstrate in a set of model host-guest systems that the recently developed Gaussian Mixture Based Enhanced Sampling (GAMBES) allows residence times to be computed and enables understanding their unbinding mechanism. We observe that guest unbinding often proceeds via a series of intermediate states that can be labelled by the number of water molecules present in the binding cavity. And in several cases the residence time is correlated to the water trapping times in the cavity.
Introduction: Beyond Binding Affinity
In drug discovery, binding affinity—how strongly a molecule sticks to its target—has traditionally been the primary metric for success. However, pharmacological efficacy is often more closely tied to "residence time," or how long a drug remains bound. Calculating these unbinding rates ($k_{off}$) is computationally demanding because molecular dynamics (MD) simulations typically operate on femtosecond scales, while unbinding events can take milliseconds or longer. Standard enhanced sampling methods often distort the physical time of the system. This study utilizes a new method, GAMBES, to bridge this gap without "breaking the clock."
Methodology: The GAMBES Approach
The authors employ Gaussian Mixture Based Enhanced Sampling (GAMBES) to accelerate rare unbinding events while preserving the ability to recover physical timescales. The process involves:
- Metastable Mapping: Identifying preferred binding poses through short, unbiased simulations.
- Gaussian Modeling: Fitting a Gaussian Mixture Model (GMM) to define the energy landscape of the bound state.
- Selective Biasing: Applying a bias potential to push the guest out, which is strategically cut off as the molecule reaches the transition region.
- Time Rescaling: Using Grubmüller/Voter rescaling to convert accelerated simulation time back to real-world time, made possible by the zero-bias condition at the transition state.
Model Systems: The SAMPL5 Challenge
The method was validated using the SAMPL5 host-guest benchmark, which provides a diverse set of chemical environments.
The study focused on two hosts: the wide-mouthed Octa-acid (OAH) and the narrow-mouthed TEMOA (OAMe). Six guests (G1-G6) varying in size, flexibility, and chemical composition were tested across both hosts, resulting in 12 unique systems.
Mechanistic Insights: Water as a Milestone
A key discovery of this work is that unbinding is not a continuous process but a "staircase" of discrete intermediate states. These states are defined by the specific number of water molecules occupying the host cavity.
In several systems, a "water scoop" mechanism was observed, where a chain of water molecules forms around the guest, physically nudging it out of the binding pocket. For the narrow-mouthed OAMe host, water molecules become trapped, creating significant energy barriers that lead to much longer residence times compared to the wide-mouthed OAH host.
Results and Discussion
The researchers successfully calculated $k_{off}$ for all 12 systems, with rates spanning four orders of magnitude ($10^2$ to $10^6$ $s^{-1}$). The unbinding events followed a Poisson distribution, confirming the physical validity of the biased sampling.
Interestingly, the study found a non-linear relationship between guest size and residence time. While steric hindrance generally increases residence time, excessively bulky guests can destabilize the binding pose, actually accelerating unbinding. Furthermore, residence times were found to correlate strongly with the trapping times of water molecules within the cavity.
Conclusion
GAMBES provides a robust framework for computing unbinding rates and uncovering the molecular mechanisms of dissociation. While host-guest systems serve as simplified models, the insights into water-mediated intermediates and steric effects have significant implications for drug design. Future work will focus on scaling this approach to more complex, flexible protein-ligand systems and automating the selection of optimal collective variables.
Summary updated