A Vaccine Designed by AI Reaches Its First Human Trial

For the first time, a vaccine whose active component was designed entirely by artificial intelligence has been given to people, and the early signs are good. Researchers at the University of Cambridge announced the milestone on 5 June, describing a small first-in-human trial of a coronavirus vaccine built around a part that no human drew up.

The thinking behind the work is a shift from how vaccines are usually made. Instead of aiming at one virus, the Cambridge team fed machine learning every available genetic sequence from the Sarbeco group of coronaviruses, the family that includes SARS-CoV-2 and the original SARS. The software then designed a "super-antigen," a synthetic protein meant to teach the immune system to recognise the whole family at once, even strains that have not appeared yet.

"We've converted vaccine development from being reactive to being future proof," said Professor Jonathan Heeney of Cambridge's Laboratory of Viral Zoonotics, who led the work. "Our vaccines will continue to provide protection against viruses even as they mutate into new strains."

The trial was deliberately modest. Thirty-nine healthy volunteers between 18 and 50 received the vaccine, delivered without a needle using a fine jet of fluid through the skin. According to the university, the shot was safe and produced no significant side effects. It also did what the designers hoped: volunteers mounted immune responses not only to SARS-CoV-2 and SARS, but to related bat coronaviruses that could one day jump to humans.

Professor Saul Faust of the University of Southampton, the trial's chief investigator, described the approach as a way to prepare for "viruses that haven't yet emerged." The work drew in Cambridge's Department of Veterinary Medicine, the spin-out company DIOSynVax, and NIHR clinical research facilities in Southampton and Cambridge.

Some caution is worth keeping. A 39-person safety study is the first rung on a long ladder, and an immune response measured in a lab is not the same as protection out in the world. The researchers plan a larger Phase 1/2 study of more than 200 volunteers, and they say the same design method is already being pointed at flu and Ebola.

What makes this notable is where the AI sat in the process. We have covered AI's growing role in biodefence and life-sciences work before, but here the model did not just sort through candidates or speed up a screen. It produced the active design that went into a human arm. That is a meaningful line to cross, and it is now crossed.