Consensus computational immunogenicity modelling of botulinum neurotoxin serotypes: Cross-platform validation,...

The recent clinical development of botulinum neurotoxin serotype E (BoNT/E), valued for its rapid onset, introduces a novel therapeutic protein with an unknown long-term immunogenic risk profile. To generate a method-agnostic risk assessment, we applied a consensus computational immunogenicity framework, rigorously calibrated against the clinically observed 1-3% neutralizing antibody (NAb) incidence for BoNT/A. Our multi-platform strategy integrated an ensemble of four independent HLA class II epitope prediction algorithms, triplicate molecular dynamics simulations using distinct force fields, and three independent systems immunology models. This triangulated approach consistently identified BoNT/E as possessing a significantly enriched epitope landscape, with 73-83% more predicted strong HLA binders than BoNT/A. Biophysical simulations confirmed that BoNT/E-derived peptides form more stable complexes with HLA molecules, exhibiting a mean ΔΔG ∼ binding ∼ advantage of approximately -13 kcal/mol. Systems-level models projected a consensus threefold increase in the hazard for NAb development (HR = 3.03) and an accelerated risk for concomitant BoNT/A + E therapy. Control analyses confirmed the specificity of the signal to native BoNT/E epitope architecture, and Bayesian modelling quantified a >99% posterior probability that BoNT/E confers higher relative immunogenic risk. These predictions remain subject to the inherent simplifications of computational models relative to the complexity of human immune responses. Nonetheless, this convergent, cross-platform evidence establishes a robust risk hypothesis, underscoring the need for enhanced clinical immunogenicity monitoring for BoNT/E.