AI and Viral Evolution: Charting the Path to Pandemic Preparedness

The ability to predict viral evolution could revolutionize public health, enabling proactive vaccine development and antiviral treatments. Recent advances in artificial intelligence (AI) offer hope for making this vision a reality, with research groups worldwide applying AI to the study of viruses like SARS-CoV-2 and influenza. By analyzing genetic sequences, AI can forecast viral mutations and variants, bringing us closer to outpacing the relentless adaptability of viruses.

AI-driven tools such as AlphaFold, ESM-2, and CoVFit are central to this effort. These models leverage vast datasets—such as the nearly 17 million SARS-CoV-2 sequences available today—to predict how viruses may evolve. For example, Harvard Medical School’s EVEscape model has been used to engineer SARS-CoV-2 spike protein variants capable of evading current antibodies, helping scientists test vaccine effectiveness preemptively. Similarly, the CoVFit model, developed at the University of Tokyo, accurately predicted the fitness of SARS-CoV-2 variants, including JN.1, which became globally dominant in 2024.

Yet, challenges persist. Sudden evolutionary leaps, like the emergence of the Omicron variant, remain difficult to predict. AI models typically focus on smaller, incremental changes, but the complexity of viral evolution demands new approaches to map these vast trajectories. Efforts to combine sequencing and experimental data offer promising avenues for refinement, providing researchers with a broader perspective on viral adaptability.

The stakes are high. As evolutionary virologist Shusuke Kawakubo explores the immune-evading potential of influenza’s haemagglutinin protein, the ultimate goal is clear: to stay ahead of viral evolution and ensure vaccines remain effective. While the work is far from over, AI’s role in this field marks a transformative step toward safeguarding global health.

References

• Smriti Mallapaty (2025). What will viruses do next? AI is helping scientists predict their evolution. Nature. Available online. doi: https://doi.org/10.1038/d41586-024-04195-3. Accessed 16 January 2025.