The Promise and Peril of Using AI to Get Ahead of the Next Pandemic

The Promise and Peril of Using AI to Get Ahead of the Next Pandemic Can Artificial Intelligence Predict the Next Viral Threat?

In recent years, researchers have made major advances in using artificial intelligence (AI) to predict viral evolution and design better vaccines and therapies. This approach holds enormous potential for preparing for future pandemics, but experts caution that the technology is still in its early stages.

The COVID-19 pandemic caught the world off guard and revealed gaps in pandemic preparedness. But thanks to previous research on coronaviruses, scientists were able to quickly develop highly effective vaccines based on the SARS-CoV-2 spike protein. "We got totally lucky, hugely lucky," says structural biologist Neil King of the University of Washington.

For many other concerning viruses like the henipaviruses, researchers have no pre-existing knowledge to guide vaccine design. This is where AI tools like the protein structure predictor AlphaFold are proving invaluable.

AlphaFold enabled researchers to map the structure of a key viral protein from Langya henipavirus, a recently identified virus in China similar to lethal Hendra and Nipah viruses. The team then used another AI to identify mutations to stabilize the protein, providing the basis for a prototype Langya vaccine. "Machine learning enabled something that would not have been possible otherwise," says structural virologist David Veesler.

Other scientists are using neural networks to generate promising protein changes for next-generation vaccines. These AI-assisted approaches "have totally changed the game in terms of what is possible," says King. Groups funded by CEPI and NIAID are compiling vaccine libraries for dangerous viruses using protein engineering powered by machine learning.

But designers still struggle to keep pace with the dizzying speed of viral evolution. For instance, COVID-19 vaccines have been perpetually several steps behind new viral variants. Each new vaccine update takes months to produce while providing limited protection.

To get ahead of shape-shifting viruses, researchers at Harvard and Oxford developed an AI system named EVEscape. It combines evolutionary predictions with viral biological details to forecast problematic mutations and rank emerging variants by threat level.

In a key test, the researchers showed EVEscape could have anticipated the most concerning SARS-CoV-2 variants if deployed early in the pandemic. "By rapidly determining the threat level of new variants, we can help inform earlier public health decisions," says lead author Sarah Gurev.

The creators of EVEscape are now using the tool to continuously rank new SARS-CoV-2 variants in real-time. This information could eventually guide development of vaccines resilient to viral escape mutations. EVEscape has also shown promise for predicting evolution of other viruses like HIV and flu.

But experts caution that AI-based prediction tools remain imperfect. "These new techniques are changing our everyday life, making things so much faster. But it’s still early days," says protein designer Clara Schoeder. The models rely on heavy processing of training data, and their recommendations require careful interpretation and validation.

Nonetheless, AI offers invaluable leads for preparedness between pandemics, when data are scarce. Researchers are applying EVEscape to understudied threats like Lassa and Nipah viruses. "We want to figure out how we can actually design vaccines and therapies that are future-proof," says Noor Youssef of Harvard Medical School.

Overall, AI has expanded possibilities for getting ahead of viruses and designing better countermeasures. "Does machine learning give us new arrows in our quivers? Yes, absolutely," affirms King. But realizing the full potential will require ongoing advances in biological insight to complement the power of modeling.

Vigilance is also key, as viruses continue evolving in unexpected ways. "We underestimate the ability of things to mutate when they’re under pressure and have a large population in which to do so," warns Debora Marks of Harvard Medical School. "Viruses are flexible—it's almost like they've evolved to evolve."