A structure-informed evolutionary model for predicting viral immune escape and evolution

Persistent emergence of viral variants capable of evading host immunity constitutes a significant threat to public health. This antigenic evolution frequently outpaces the development of vaccines and therapeutics, highlighting the necessity of predictive models forecasting the immune escape potential of emerging variants. However, existing models suffer from two key limitations: they inadequately incorporate protein structural information and neglect the importance of distinguishing large-impact mutations from neutral ones given multiple mutations. To address these gaps, we presented EvoScape, a deep learning model designed to predict viral immune escape and evolution by integrating evolutionary context with protein structure, capturing residue-residue dependence, and introducing a novel top-K L₂-differential pooling mechanism to prioritize mutations with large functional impacts. We demonstrated that EvoScape significantly outperformed state-of-the-art models on the most comprehensive benchmark to date, comprising eleven deep mutational scanning experiments spanning diverse viruses. Furthermore, EvoScape exhibited outstanding performance in the practical applications of identifying immune escape hotspots and forecasting circulating strains of SARS-CoV-2. These results show that EvoScape is a powerful model to predict viral immune escape and evolution. Its capacity for early warning can directly inform public health interventions and guide the development of countermeasures, thereby mitigating the threat of future viral pandemics.