Deep Mutational Scanning and Computational Modeling of SARS-CoV-2 Spike Protein Receptor-Binding Domain Escape from Neutralizing Antibodies
Introduction
The emergence of SARS-CoV-2, a betacoronavirus with documented spillover into multiple mammalian species including mink, white-tailed deer, felids, and mustelids, has underscored the necessity of understanding antibody escape mechanisms at the molecular level in a veterinary context [1, 2]. The spike glycoprotein, specifically its receptor-binding domain (RBD), mediates attachment to angiotensin-converting enzyme 2 (ACE2) and constitutes the primary target of neutralizing antibodies (nAbs) [3]. Continuous antigenic drift in the RDB drives immune evasion, compromising both natural immunity and diagnostic serological assays used in animal health surveillance [4]. Deep mutational scanning (DMS) combined with computational structural biology provides a systematic framework to quantify the impact of individual amino acid substitutions on antibody binding and viral fitness [5, 6]. This article reviews the integration of DMS experiments, structure-based free energy calculations, machine learning models, and large-scale sequence surveillance to predict and characterize escape mutations in the SARS-CoV-2 spike RDB.
Principles of Deep Mutational Scanning
Deep mutational scanning is a high-throughput mutagenesis approach that measures the functional consequences of thousands of single amino acid substitutions in a protein of interest [1]. For the SARS-CoV-2 spike RDB, libraries of mutant variants are expressed on the surface of yeast or mammalian cells, exposed to a monoclonal antibody or polyclonal serum, and then sorted by flow cytometry to separate bound (neutralized) from unbound (escape) populations [2, 7]. Deep sequencing of the input and output populations yields enrichment ratios that quantify the relative fitness of each mutant under antibody pressure [8]. The resulting escape maps highlight residues where substitutions reduce antibody binding, thereby conferring immune evasion [9].
Epistatic interactions, in which the effect of a mutation depends on the genetic background, have been shown to modulate escape landscapes across emerging SARS-CoV-2 lineages [1]. Taylor and colleagues demonstrated that amino acid preferences at key RBD positions shift in the context of Omicron subvariants, reflecting epistatic rewiring of the fitness landscape [1]. Similarly, Shao et al. applied DMS to JN.1 and XEC RBDs to reveal lineage-specific escape profiles against a panel of monoclonal antibodies [2]. These studies illustrate that DMS data must be interpreted within the evolutionary context of circulating variants.
Computational Modeling of Mutation Effects on Antibody Binding
Structure-based computational methods predict the biophysical impact of RBD mutations on antibody binding affinity and thereby complement DMS data [4, 7]. The binding free energy change (ΔΔG) upon mutation can be estimated using Rosetta, FoldX, or molecular dynamics (MD) simulations [10]. Kunkel et al. employed all-atom MD simulations to characterize spike protein dynamics, demonstrating that mutations at the RBD interface remodel the epitope landscape and alter antibody recognition [11]. Durumeric and colleagues integrated DMS measurements with machine learning-driven simulations to generate continuous fitness landscapes for the RBD, enabling prediction of escape mutations that confer high fitness and low antibody binding [4].
AlphaFold2 and related deep learning architectures have been used to predict the three-dimensional structures of variant RBDs and their complexes with antibodies [3]. By docking the mutant RBD onto a representative antibody structure, one can compute interface buried surface area, hydrogen bond networks, and van der Waals contacts to rationalize escape [7]. Soliman et al. systematically analyzed RBD evolution from ACE2 binding optimization to immune epitope remodeling, highlighting how mutations at the receptor-binding motif (RBM) simultaneously modulate receptor affinity and antibody evasion [7]. The integration of structural modeling with DMS allows researchers to prioritize mutations for experimental validation and to anticipate future escape variants.
Machine Learning and Protein Language Models for Predicting Escape
Recent advances in machine learning have leveraged DMS data to train predictive models of viral evolution [3, 5, 6]. Yang and colleagues developed a DMS-informed protein language model that captures spatiotemporal evolutionary dynamics of SARS-CoV-2, accurately forecasting the emergence of antibody-resistant lineages [3]. The model uses embeddings from pre-trained language models (e.g., ESM-1v) and fine-tunes them on DMS enrichment scores to predict the effect of unseen mutations on antibody escape and receptor binding [10].
Nasir et al. applied machine learning classifiers to group SARS-CoV-2 variants based on antigenic similarity, using DMS-derived escape data as input features [5]. This approach enables real-time antigenic cartography similar to that used for influenza viruses, allowing veterinary diagnosticians to assess whether new animal isolates are likely to escape existing immune pressure. Shlesinger and colleagues developed a deep mutational learning framework that deconvolves polyclonal serum escape into contributions from individual antibody specificities [6]. By training neural networks on DMS data from convalescent sera, the model quantifies how each RBD mutation reduces neutralization breadth, providing a composite escape score for emerging variants.
Lamb et al. demonstrated that protein language models trained on evolutionary sequence data can predict the mutational potential of the spike protein without explicit DMS measurements [10]. When combined with DMS data, these models produce robust forecasts of which mutations are likely to become fixed under antibody selection. The synergy between DMS and language models represents a powerful paradigm for anticipating antigenic drift in both human and animal coronaviruses.
Integration with Large-Scale Sequence Surveillance
Escape mutations identified through DMS and computational modeling must be monitored in real-world sequence data to assess public health and veterinary risk. Platforms such as GISAID (Global Initiative on Sharing All Influenza Data) curate millions of SARS-CoV-2 genome sequences from humans and animals, enabling the detection of emerging variants carrying predicted escape mutations [2, 9]. Haddox and colleagues examined clonal interference and shifting selective pressures on the RBD across the pandemic, demonstrating that antibody escape mutations accumulate concurrently with ACE2 affinity-enhancing substitutions [9]. Their work underscores the importance of considering both immune pressure and receptor binding efficiency when evaluating variant fitness in animal hosts.
The identified escape hotspots, such as positions 346, 356, 440, and 484, are recurrent sites of substitution in Omicron sublineages and have been observed in mink and deer isolates [1, 2]. Veterinary surveillance programs can incorporate computational escape predictions by screening animal-derived sequences for these hotspots [7]. Ding and Yuan developed a mathematical model that integrates receptor binding and immune escape to predict the fitness landscape of SARS-CoV-2 variants, validating their predictions against global sequence data [8]. Such models are directly transferable to animal coronaviruses where similar DMS datasets become available.
Table 1: Computational Methods for Predicting Antibody Escape in the SARS-CoV-2 Spike RBD
| Method | Approach | Output | Key References |
|---|---|---|---|
| Deep Mutational Scanning (DMS) | High-throughput mutagenesis + deep sequencing | Enrichment scores per mutation | [1, 2, 9] |
| Structure-based energy calculations (Rosetta, FoldX) | ΔΔG prediction from atomic models | Binding free energy change | [4, 7, 11] |
| Molecular dynamics simulations | All-atom trajectories of RBD-antibody complex | Conformational dynamics, interface stability | [11] |
| Protein language models (e.g., ESM) | Evolutionary sequence embeddings + fine-tuning | Mutational effect prediction | [3, 10] |
| Machine learning classifiers | Training on DMS data to predict antigenic group | Antigenic cartography | [5, 6] |
| Fitness landscape simulations | Integration of DMS and deep mutational learning | Continuous landscape of escape + fitness | [4, 8] |
Mermaid Workflow: Integrated Pipeline for Predicting Antibody Escape
graph TD
A[Generate RBD Mutant Library], > B[Deep Mutational Scanning Flow Cytometry]
B, > C[Deep Sequencing & Enrichment Scores]
C, > D[Escape Map: Residue-level antibody sensitivity]
D, > E[Structure-based modeling: ΔΔG, docking, MD]
D, > F[Machine learning: language models, classifiers]
E, > G[Predicted escape mutations + fitness scores]
F, > G
G, > H[Comparison with GISAID animal/human sequences]
H, > I[Identify emerging escape variants in animal hosts]
I, > J[Update diagnostic assays & vaccine strains for veterinary use]
style A fill:#f9f,stroke:#333,stroke-width:2px
style J fill:#bbf,stroke:#333,stroke-width:2px
Veterinary Implications and Comparative Perspectives
While the DMS and computational modeling methods described above were developed predominantly for SARS-CoV-2, their principles are directly applicable to animal coronaviruses, including canine respiratory coronavirus, feline coronavirus, and mink-derived variants. The ACE2 receptor is highly conserved among mammals, and the structural interface of the RBD with ACE2 shows substantial homology [7]. Reverse zoonotic transmission events have introduced human SARS-CoV-2 variants into farmed mink and wild deer, where subsequent evolution has been observed [2]. Monitoring antibody escape in these animal reservoirs requires an integrated computational pipeline that predicts which RBD mutations would undermine serological detection or vaccine-induced immunity in veterinary species.
The computational approach also informs the design of monoclonal antibody therapies for veterinary use. By mapping escape mutations in silico, one can select antibody cocktails that target conserved epitopes less prone to mutational escape [6, 9]. Furthermore, protein language models trained on coronavirus sequence data from multiple hosts can predict host-specific adaptation and immune evasion, aiding risk assessment for future spillover events [3, 10]. The veterinary community benefits from adopting these bioinformatic tools to complement traditional surveillance and diagnostics.
Conclusion
Deep mutational scanning provides empirical measurements of how single amino acid substitutions in the SARS-CoV-2 spike RBD affect neutralization by antibodies. When combined with computational structural modeling, molecular dynamics simulations, and machine learning, DMS data yield predictive models of viral escape that are generalizable to animal coronaviruses. The integration of these approaches with global sequence surveillance enables proactive identification of emerging variants in both human and animal populations. As DMS datasets expand to include additional coronaviruses, the computational frameworks reviewed here will become essential for veterinary virology, diagnostic assay development, and pandemic preparedness.
References
[1] Taylor AL, Starr TN. Deep mutational scanning of recent SARS-CoV-2 variants highlights changing amino acid preferences within epistatic hotspot residues. PLoS Pathog. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42330076/
[2] Shao C, Yang L, Xiao C, et al. Deep mutational scanning reveals the antibody escape and infectivity landscape of SARS-CoV-2 Omicron JN.1 and XEC receptor-binding domains. Emerg Microbes Infect. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42324717/
[3] Yang S, Luo X, Luo J, et al. A deep mutational scanning-informed protein language model predicts SARS-CoV-2 evolution dynamics with spatiotemporal resolution. Nat Microbiol. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42204343/
[4] Durumeric AEP, McCarty S, Smith J, et al. Machine Learning-Driven Simulations of the SARS-CoV-2 Fitness Landscape from Deep Mutational Scanning Experiments. J Chem Inf Model. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42089465/
[5] Nasir A, Lee D, Avena LE, et al. Predictive modeling of immune escape and antigenic grouping of SARS-CoV-2 variants. J Virol. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42037411/
[6] Shlesinger D, Sadilek V, Minot M, et al. Dissecting serum polyclonal antibody escape to SARS-CoV-2 variants by deep mutational learning. Cell Rep Methods. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42030951/
[7] Soliman OA, Shahine Y, Baecker D, et al. Beyond the Mutation Abyss: Revisiting SARS-CoV-2 Receptor-Binding Domain Evolution from ACE2 Binding Optimization to Immune Epitope Remodeling. Pathogens. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41901725/
[8] Ding Z, Yuan HY. The role of receptor binding and immunity in SARS-CoV-2 fitness landscape: A modeling study. iScience. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41809055/
[9] Haddox HK, Abdel Aziz O, Galloway JG, et al. Clonal interference and changing selective pressures shape the escape of SARS-CoV-2 from hundreds of antibodies. Virus Evol. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41767406/
[10] Lamb KD, Hughes J, Lytras S, et al. From single-sequences to evolutionary trajectories: protein language models capture the evolutionary potential of SARS-CoV-2. Nat Commun. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41714330/
[11] Kunkel G, Madani M, White SJ, et al. Modeling coronavirus spike protein dynamics: implications for immunogenicity and immune escape. Biophys J. 2021. URL: https://pubmed.ncbi.nlm.nih.gov/34767789/ *** 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.