Variant-aware computational assessment of neutralizing antibody binding robustness to botulinum toxin A

Botulinum neurotoxin type A (BoNT/A) is a widely used biologic therapeutic, yet the influence of natural toxin sequence variation on the stability of neutralizing antibody binding remains poorly understood. In this study, we performed a comprehensive computational analysis to quantitatively assess antibody binding robustness across naturally occurring BoNT/A variants. Using a curated dataset spanning all major subtypes, we integrated sequence variability mapping, ensemble antibody docking, explicit-solvent molecular dynamics simulations (totalling >30 μs), Bayesian hierarchical modelling, and interface network analysis. Sequence analysis revealed that amino acid variability is strongly enriched in solvent-exposed regions, with mean Shannon entropy increasing from 0.07 in the light-chain core to 0.48 in surface loops of the heavy chain, and a significant association between entropy and solvent accessibility (Wilcoxon p < 10^-6). Docking showed modest variant-dependent energetic dispersion (Cohen's d < 0.5) that poorly correlated with dynamic stability metrics (ρ = 0.18-0.31). In contrast, molecular dynamics simulations identified large variant-specific effects on interface stability, with multiple variants exhibiting >1 standard deviation increases in interface root-mean-square deviation (RMSD, a measure of average atomic displacement between structures) and reductions in contact persistence and hydrogen-bond occupancy. A composite Binding Robustness Index (BRI) separated high- and low-robustness variants by 2.1 to 2.7 standard deviations within antibody systems. Bayesian modelling indicated that 31-46% of variants had high posterior probability (>0.95) for reduced binding robustness. Interface network analysis revealed that low-robustness variants consistently disrupted cooperative interface topology, with 35-61% loss of high-betweenness residues and increased modularity, strongly correlating with Bayesian effect estimates (ρ = 0.74). These results demonstrate that natural BoNT/A sequence variation leads to substantial, antibody-specific differences in binding robustness, driven by localized disruption of cooperative interface networks rather than uniform binding weakening. This study provides a quantitative computational framework for assessing antibody binding stability across toxin variants, with implications for therapeutic durability and resistance mechanisms.