世界上第一个AI设计的疫苗解析

Researchers at the University of Cambridge have developed what they describe as a fundamentally new type of vaccine using artificial intelligence (AI) . The vaccine’s key component was designed entirely by AI and has now been tested in people for the first time.

The goal is ambitious: a single vaccine that works not just against all known human coronavirus variants, but against related bat viruses that could jump from animals to humans and cause future pandemics.

Traditional vaccines train our immune system to recognise one specific virus. The problem is that viruses mutate. When they change enough, the vaccine stops working, which is why we need a new flu shot every year and why COVID vaccines have been updated repeatedly since 2021.

AI offers a way around this. By analysing genetic data from thousands of related viruses, it can identify the parts that stay the same across different strains and that are unlikely to change over time. Target those stable features, and you have a vaccine that should work against the whole family, not just the strain you started with.

This is exactly what the Cambridge team did. They used AI to scan viruses from the sarbecovirus family, which includes the viruses that cause both SARS and COVID, as well as a range of animal coronaviruses – looking for shared features that evolution has left largely untouched. Those features became the basis of the vaccine.

DNA vaccines

While many people are familiar with the mRNA shots used during the pandemic, this new vaccine uses DNA. DNA vaccines are generally more stable than mRNA vaccines, making them easier to store and transport. A significant advantage in lower-income countries where “cold-chain” infrastructure is limited.

They can also be administered without needles. A high-pressure stream of liquid delivers the vaccine through the skin, making administration less painful and easier to scale up during an outbreak.

Could it protect against future pandemics?

These practical advantages matter most if the vaccine itself can do something no existing jab can: protect against viruses we haven’t encountered yet.

Broad-spectrum vaccines could change the way the world responds to emerging infectious diseases. By offering much wider protection than traditional vaccines, they could provide rapid immunity against new and emerging viral threats. This would equip public health officials with tools to stop future outbreaks in their tracks before they have a chance to turn into global pandemics.

They could also transform our approach to more familiar diseases. Influenza is a prime target because it exists in many different strains and evolves so rapidly. Scientists have to predict which strains will dominate each flu season, and they guess wrong, vaccine effectiveness can suffer. A universal flu vaccine that targets features shared across multiple strains could eventually end the annual race to keep up with the virus.

And the Ebola virus shows why this matters right now. The recent outbreak in the Democratic Republic of the Congo and Uganda is driven by the Bundibugyo strain, which bypasses existing vaccines. While researchers rush to create a new vaccine specifically for this strain, local communities remain at high risk. A broad-spectrum vaccine designed to cover an entire virus family could transform that picture.

What the trial found

This is the first human trial of an AI-designed vaccine. The results showed that this DNA vaccine was able to stimulate the immune system to produce antibodies that can recognise different types of sarbecoviruses. The technology was found to be safe and well tolerated.

This is an exciting advance because it demonstrates how AI has the potential to design variant-proof vaccines against future pandemic threats. The needle-free delivery system could also make the vaccine easier to administer and distribute worldwide.

However, there is more work to do. Although the results in this study are encouraging, the immune responses following vaccination were modest. It was also uncertain how long the protection lasts and whether further boosters will be required. Larger trials are also needed to determine whether the vaccine can prevent or reduce virus infections in the real world.

A universal vaccine remains a few years away. And any new vaccine must still pass larger trials to prove it is safe, effective and provides lasting protection. But this study shows the goal is getting closer – and AI may help us get there faster.

Neil Mabbott currently receives funding from the UK Biotechnology and Biological Research Council, The Creutzfeldt-Jakob Disease Foundation, and Zoetis.