Toward Robust Characterization of Dynamic Binding Pockets: Lessons from the HBV Capsid Assembly Modulator Site

Technical Summary: Toward Robust Characterization of Dynamic Binding Pockets

Problem Statement
Protein function is often dictated by ligand binding pockets that fluctuate among conformational states, altering their size, shape, topology, and accessibility. Quantitative comparison of these dynamic cavities remains challenging because their boundaries are inherently ambiguous, particularly when comparing apo and ligand-bound structures or ensembles derived from molecular dynamics (MD) simulations. Existing computational methods (grid-based, probe-based, rolling-ball, etc.) rely on how the molecular boundary is defined; variations in this definition can significantly alter calculated volumes and geometries, complicating cross-study comparisons. Furthermore, parameters optimized for static experimental structures often fail to remain robust across the broader conformational landscapes of dynamic proteins or lack the efficiency required for high-throughput analysis. While the measure volinterior algorithm in Visual Molecular Dynamics (VMD) offers a framework for characterizing enclosed spaces using fuzzy-boundary detection, practical guidance for adapting this method to small, partially solvent-accessible binding pockets—rather than large molecular containers like virus capsids—has been lacking.

Methodology
The authors utilize the Hepatitis B Virus (HBV) capsid assembly modulator (CAM) binding site as a model system to develop and validate a practical workflow for applying measure volinterior to dynamic pockets. The CAM site is an ideal test case due to its substantial conformational heterogeneity, its formation at a dynamic protein–protein interface, and its sensitivity to boundary definitions.

The methodology focuses on optimizing three interdependent parameter components:

1. Definition of the Molecular Container: Instead of using the entire protein, the workflow defines a localized subset of residues. The authors compare "ligand-guided" approaches (selecting residues within a distance cutoff of a bound ligand) against "pocket-guided" approaches (defining the container based on protein architecture). They demonstrate that ligand-guided definitions are sensitive to ligand identity, binding mode, and occupancy, whereas a pocket-guided approach, refined iteratively through visual inspection and relative Solvent Accessible Surface Area (SASA) analysis, provides a ligand-independent definition robust across apo and bound states.
2. Construction of the Molecular Surface: The study optimizes the QuickSurf representation parameters (Radius Scale, Isovalue, and Grid Spacing) and the Number of Rays. The authors emphasize that these parameters must be tuned together to balance surface smoothness, compactness, and spatial resolution. They specifically recommend using the GPU implementation of measure volinterior with VMD's default van der Waals radii, noting that CPU and GPU implementations yield different results.
3. Selection of the Occlusion Threshold: Using fuzzy-boundary detection, the authors determine the effective threshold (ranging from 0.1 to 0.9) that defines the pocket boundary across its solvent-accessible opening. They evaluate how different thresholds affect the continuity and physical relevance of the detected cavity.

The workflow includes strategies for visualizing results (using isosurfaces or dummy-atom point clouds) and addresses systematic errors introduced by rigid-body rotations, recommending alignment to a common reference frame to minimize voxel-grid artifacts.

Key Results

• Optimized Protocol: A standardized parameter set for the HBV CAM site was established (Table 1): Atom Selection based on specific base and capping chain residues; Radius Scale of 1.2 Å; Isovalue of 0.8; Grid Spacing of 0.5 Å; Number of Rays of 64; and an Occlusion Threshold of 0.9.
• Quasi-Equivalence and Dynamics: Application of this protocol to four crystal structures (apo, AT130-bound, HAP1-bound, HAP18-bound) revealed pronounced differences in pocket volume among the four quasi-equivalent sites (A, B, C, D), even in the apo form. For instance, the B-site was found to be 177% the size of the A-site in the apo structure, suggesting quasi-equivalence significantly influences geometry independent of ligand binding.
• Robustness Across Ensembles: The optimized parameters were validated against a 1-µs all-atom MD simulation containing three million conformations. The protocol successfully maintained a localized, continuous cavity representation across extreme conformational states (highly contracted and expanded pockets), whereas ligand-guided definitions would have failed for apo or partially resolved structures.
• Systematic Errors: The study quantified orientation-dependent systematic errors, finding that rigid-body rotations could introduce volume variations of up to ~70 ų. This highlights the necessity of aligning structures to a common reference before quantitative comparison.
• Visualization: The use of dummy-atom point clouds was shown to be a more faithful and efficient representation of small cavity geometries compared to pre-equilibrated water boxes, which can introduce artifacts due to the discrete nature of water molecules.

Significance and Claims
The paper claims to provide a reproducible strategy for adapting measure volinterior to other dynamic binding pockets, moving beyond its original application to large molecular containers. By establishing a standardized protocol for the HBV CAM site, the authors enable consistent, quantitative comparison of pocket geometry across independent structural studies and MD ensembles.

The authors assert that the primary contribution is not a universal set of "optimal" numbers, but rather an iterative, data-driven workflow for developing robust parameterizations. They argue that successful characterization requires parameters that:

1. Remain localized to the pocket of interest without detecting unrelated cavities.
2. Preserve continuous topology where the protein architecture allows.
3. Approximate the steric boundary without encroaching on the pocket or protein volume.
4. Reflect genuine conformational rearrangements rather than parameter artifacts.

The work provides practical starting ranges for parameters (e.g., Radius Scale 1.0–1.2 Å, Occlusion Threshold 0.7–0.9) and emphasizes that parameter selection must be guided by the resulting cavity geometry and physical interpretability rather than numerical maximization or minimization of volume. This approach aims to facilitate structure-based drug design and protein engineering by providing reliable, reproducible metrics for dynamic molecular recognition.