AI-based platform for predicting emerging vaccine-escape variants and designing mutation-proof antibodies

This project, led through a collaborative effort focusing on an AI-based platform to predict emerging vaccine-escape viral variants and to guide the development of evolution-resistant antibodies and vaccines. With a $2.7 million award from the National Institutes of Health, the team is advancing machine learning and mathematical models that combine global viral genomic data with experimental virology to forecast how viruses mutate and which changes will materially affect infectivity, immune evasion and therapeutic effectiveness. The lab seeks to develop safe surrogate systems, pseudo-viruses and lab-grown cell assays, and highly sensitive tests that reveal which viral variants infect which cells, enabling rapid experimental confirmation of computational forecasts.

The platform's predictive capabilities were stress-tested during the COVID-19 pandemic. Using live genomic data and their AI framework, the team identified key spike protein sites whose mutation later proved central to prevalent SARS-CoV-2 variants. When the first omicron variant emerged in late 2021, Wei and Zheng's models rapidly projected greater infectivity, enhanced immune evasion and reduced responsiveness to existing antibody therapies, predictions that were later borne out by experimental work worldwide. Their forecasts offered a window of weeks to months for public health planning and therapeutic adjustment. A similar sequence occurred with the BA.2 omicron subvariant in early 2022, where the team anticipated increased transmissibility and immune escape that aligned with subsequent global observations.

Beyond immediate pandemic response, the project aims to generalize the platform across viral families - influenza, HIV, Ebola and others - leveraging shared evolutionary principles to design broadly protective, "evolution-proof"  vaccines and therapies. The AI models are being refined to improve timeliness and accuracy so they can inform public health decisions, help developers prioritize vaccine updates or antibody candidates, and suggest targets less likely to accommodate escape mutations. Critically, the research integrates computation and empirical validation: every in silico prediction is tested in biological systems to ensure real-world relevance.

By strengthening interdisciplinary collaboration across mathematics, computational science and experimental virology, the effort seeks to shift reactive pandemic responses toward anticipatory preparedness. Ultimately, the project aspires to provide reliable forecasts of viral evolution that enable developers and public health officials to design and deploy vaccines and antibody therapies that remain effective in the face of viral change.