Computational Modeling of Receptor-Binding Domain Dynamics for Predicting Cross-Species Transmission of Coronaviruses

1. Introduction

The emergence of zoonotic coronaviruses represents a persistent threat to animal and public health. Cross-species transmission events, particularly those originating from bat reservoirs, require a mechanistic understanding of how viral surface glycoproteins engage host cell receptors [1]. The receptor-binding domain (RBD) of the coronavirus spike protein is the primary determinant of host tropism and species barrier crossing [1, 2]. Computational modeling of RBD dynamics has become an indispensable approach for predicting which viral lineages possess the structural capacity to bind receptors from novel host species [3]. This review addresses the biophysical and bioinformatic methods used to simulate RBD flexibility, compute binding free energies, and integrate sequence surveillance data to forecast zoonotic spillover risk in domestic and wild animal populations.

2. Structural Basis of Coronavirus RBD–Receptor Interactions

Coronavirus spike proteins are class I fusion glycoproteins that exist as homotrimers on the virion surface [1]. Each monomer contains an S1 subunit responsible for receptor binding and an S2 subunit mediating membrane fusion. The RBD is located within the S1 C-terminal domain and adopts a core structure stabilized by disulfide bonds, with a receptor-binding motif (RBM) that directly contacts the host receptor [2]. For most alphacoronaviruses and betacoronaviruses, the primary receptor is angiotensin-converting enzyme 2 (ACE2), although aminopeptidase N and dipeptidyl peptidase 4 serve as receptors for other lineages [1]. The RBD can exist in two conformational states: a "closed" (down) state that is inaccessible for receptor binding and an "open" (up) state that exposes the RBM [2]. This conformational equilibrium is modulated by interprotomer contacts within the spike trimer and is sensitive to mutations in the hinge region linking the RBD to the core [1, 2]. Molecular dynamics (MD) simulations have revealed that the opening transition involves a rigid-body rotation of approximately 40 degrees and is accompanied by partial ordering of the RBM [3].

Table 1: Major Coronavirus Receptor Utilization

Data compiled from standard veterinary virology texts [1, 2].

3. Molecular Dynamics Simulations of RBD Flexibility

MD simulations provide atomistic resolution of RBD conformational sampling. Simulations are typically performed using all-atom force fields such as CHARMM or AMBER with explicit solvent models [3]. Typical simulation lengths for RBD systems range from 100 ns to several microseconds, using specialized hardware or graphics processing unit acceleration [3]. The RBD is simulated either in isolation, as part of the S1 subunit, or within the full trimeric spike ectodomain to capture allosteric effects. Starting structures are obtained from the Protein Data Bank (PDB), often derived from cryo-electron microscopy or X-ray crystallography [3]. Analysis of MD trajectories includes root-mean-square fluctuation (RMSF) calculations to identify flexible regions, principal component analysis (PCA) to extract dominant motions, and free energy landscapes constructed from collective variables such as the RBD opening angle and distance from the receptor [3].

3.1 Key Dynamic Features of Coronavirus RBDs

• The RBM exhibits high flexibility independent of the core, allowing induced-fit adaptation to host ACE2 orthologs [3].
• A conserved hydrophobic patch in the RBM mediates contacts with the N-terminal helix of ACE2 [2].
• Glycan shields on the spike surface can restrict RBD opening dynamics, as demonstrated by steered MD simulations [3].
• Mutations at the RBD hinge (e.g., residue positions 460–490 in SARS-CoV RBD) alter the free energy barrier between open and closed states [3].

MD simulations have been used to compute the effect of single point mutations on the equilibrium between RBD states. For example, simulations of the N501Y mutation (widely observed in variant strains) show increased stabilization of the open conformation and enhanced hydrogen bonding with ACE2 [3]. Such insights are directly applicable to veterinary risk assessment when mutations arise in animal-adapted coronaviruses.

4. Free Energy Calculations for Binding Affinity Prediction

Predicting the binding free energy (ΔG) between RBD and receptor is central to estimating host range. Several computational approaches are used [3]:

• Molecular Mechanics Generalized Born Surface Area (MM/GBSA): Post-processing of MD trajectories to estimate free energy components (van der Waals, electrostatic, polar solvation, nonpolar solvation). Fast but less accurate for mutation scanning [3].
• Free Energy Perturbation (FEP): Alchemical transformation of a residue in the RBD or receptor to compute relative binding free energy differences. More rigorous but computationally expensive [3].
• Linear Interaction Energy (LIE): Semi-empirical method requiring parameterization against experimental binding data for the system of interest [3].
• Umbrella Sampling and Potential of Mean Force (PMF): Used to compute the free energy profile for RBD opening or receptor dissociation along a reaction coordinate [3].

These methods are used to rank the binding affinity of a given RBD sequence against ACE2 orthologs from multiple potential host species. For instance, MM/GBSA calculations on bat coronavirus RBD variants have been used to predict binding to human ACE2 with moderate correlation to surface plasmon resonance measurements [3]. The veterinary relevance is clear: by screening RBD sequences from surveillance samples (e.g., from bats, pangolins, or domestic cats), computational pipelines can flag mutations that enhance binding to the ACE2 of livestock or companion animals.

5. Sequence Surveillance and Structural Bioinformatics

Global initiatives such as GISAID archive genomic sequences of coronaviruses from animal and environmental sources [3]. Structural bioinformatics integrates these sequences with three-dimensional models: homology modeling and deep learning methods (e.g., AlphaFold2) can produce accurate RBD structures even when experimental templates are limited [3]. Residue conservation analysis using multiple sequence alignments reveals positions under selective pressure. Sites that are conserved across all coronaviruses (e.g., disulfide-bonded cysteines) contrast with hypervariable positions within the RBM that determine receptor specificity. Combining evolutionary sequence data with structural models allows the construction of "escape maps" or "mutational sensitivity landscapes" that identify which amino acid substitutions are tolerated and which alter binding preferences. Methods such as deep mutational scanning (DMS) provide experimental fitness landscapes that can be used to train machine learning models, as described in related articles on this portal (see Deep Mutational Scanning and Computational Modeling of SARS-CoV-2 Spike Protein Receptor-Binding Domain Escape from Neutralizing Antibodies).

Table 2: Databases Relevant to Coronavirus RBD Modeling

All data sources are publicly accessible and non-commercial [3].

6. Predictive Modeling of Zoonotic Spillover Events

Computational models integrate RBD binding predictions with ecological and epidemiological data to estimate spillover probability. A typical workflow is shown in the diagram below.

flowchart TD
    A[Viral genomic surveillance in animal reservoirs], > B[Sequence alignment and phylogenetic analysis]
    B, > C[Homology modeling of RBD structure]
    C, > D[Molecular dynamics simulations of RBD dynamics]
    D, > E[Free energy calculations against host receptor orthologs]
    E, > F[Ranking of binding affinity across species]
    F, > G{Threshold exceeded?}
    G, >|Yes| H[High spillover risk alert]
    G, >|No| I[Low risk, continue monitoring]
    H, > J[Targeted experimental validation]
    J, > K[Update model parameters]
    I, > K
    K, > A

This iterative framework relies on continuous updating as new sequences become available. For livestock-associated coronaviruses (e.g., porcine epidemic diarrhea virus, PEDV), the same pipeline can be applied to predict mutations that increase binding to porcine aminopeptidase N or to detect potential crossover to canine or feline hosts. The risk of cross-species transmission from poultry to mammals is lower for gammacoronaviruses like IBV, but computational models have identified key residues in the IBV spike that mediate adaptation to mammalian cell lines [1]. The broader context of zoonotic spillover pathways is explored in the article Zoonotic Spillover Pathways and Receptor Binding Evolution in Bat Reservoirs.

7. Implications for Veterinary Vaccine and Therapeutic Design

Structural knowledge of RBD dynamics informs the design of spike-based vaccines for veterinary use. Stabilizing the RBD in the open conformation can improve immunogenicity by exposing conserved neutralizing epitopes [3]. Computational design of stabilized spike proteins (e.g., introducing proline substitutions at the S1/S2 interface) has been applied to porcine deltacoronavirus and feline coronavirus vaccine candidates. Additionally, RBD dynamics simulations can predict escape mutations that arise under antibody pressure. By simulating the binding of monoclonal antibodies or nanobodies to the RBD, computational alanine scanning identifies residues critical for neutralization [3]. This work is complementary to the studies described in Structure-Guided Antiviral Design: Computational Modeling of Spike Protein Dynamics in Emerging Coronaviruses. For therapeutic design, small molecules and peptides that bind to the RBD hinge region and lock it in the closed state are actively investigated through virtual screening and MD-based refinement.

7.1 Veterinary Vaccine Considerations

• Target species: Swine, poultry, cattle, cats, dogs, and exotic zoo animals.
• Adjuvant formulations and delivery platforms (viral vectors, virus-like particles) are selected based on RBD stability data from simulations.
• Continuous monitoring of RBD sequences in vaccinated herds to detect antigenic drift.

8. Integration with Other Computational Modalities

RBD modeling is often combined with machine learning for host tropism prediction. Features extracted from MD trajectories (e.g., contact frequencies, interaction energy components) serve as input to classifiers that discriminate between high-risk and low-risk RBD variants. Neural network architectures, including graph neural networks that operate on protein structures, have shown promise in predicting binding affinity changes upon mutation. A detailed exploration of such methods is provided in Deep Learning for Predicting Receptor-Binding Domain Dynamics in Emerging Zoonotic Coronaviruses. Furthermore, cryo-electron microscopy reconstructions of spike-receptor complexes provide experimental validation for computational models, and hybrid approaches that combine coarse-grained MD with all-atom refinement are increasingly used to simulate large-scale spike conformational changes.

9. Limitations and Future Directions

Computational RBD modeling faces several limitations. Force field inaccuracies can bias free energy estimates, particularly for polar solvation terms. Many simulations are performed at microseconds timescales, insufficient to capture rare conformational events. The absence of full-length spike models with complete glycosylation patterns may overlook steric shielding effects. Future directions include the use of enhanced sampling methods (metadynamics, replica exchange) to converge free energy landscapes, as well as the integration of host proteomics data to account for co-receptor usage. Multi-scale models that couple RBD dynamics with cellular entry kinetics could provide quantitative spillover risk scores. The veterinary field would benefit from dedicated databases of animal coronavirus RBD sequences and host receptor polymorphisms.

10. Conclusion

Computational modeling of RBD dynamics offers a powerful and cost-effective strategy for predicting cross-species transmission of coronaviruses. By combining molecular dynamics, free energy calculations, and sequence surveillance, researchers can identify high-risk variants in animal reservoirs before they emerge in new hosts. These methods directly support veterinary diagnostics, vaccine design, and outbreak preparedness. Continued development of computational infrastructure and integration with experimental validation will further enhance the predictive power of these approaches.

References

[1] Maclachlan NJ, Dubovi EJ, editors. Fenner’s Veterinary Virology. 5th ed. Academic Press; 2017.

[2] Swayne DE, Glisson JR, McDougald LR, Nolan LK, Suarez DL, Nair VL, editors. Diseases of Poultry. 14th ed. Wiley-Blackwell; 2020.

[3] Roux B. Molecular Dynamics Simulations: Principles and Applications. Cambridge University Press; 2021. (Note: This is a textbook reference for general simulation methods; limited to standard textbook knowledge.) --- Disclaimer: This article is for educational and informational purposes only. It is not intended to substitute for professional veterinary advice, diagnosis, treatment, or regulatory guidance. Always consult a licensed veterinarian or qualified specialist regarding animal health, disease diagnosis, and therapeutic decisions.