For the first time, a vaccine whose active ingredient was designed from scratch by artificial intelligence has cleared a human safety trial — and it's aimed at every coronavirus at once. Researchers reported on June 5 that pEVAC-PS, an AI-designed universal coronavirus vaccine, was safe, well tolerated, and able to trigger broad immune responses in a Phase 1 study run at a Cambridge laboratory.
The trial, detailed in a peer-reviewed paper, enrolled 39 healthy adults across four dose levels of the needle-free, nasal-spray candidate. No serious side effects were reported at any of the doses, and participants generated antibodies and T-cell responses that recognized multiple coronavirus strains, including SARS-CoV-2 and several of its variants.
Designed by machine, tested in people
What makes pEVAC-PS unusual is how it was created. Rather than tweaking an existing vaccine target, the research team used AI models to predict which protein fragments — short stretches called peptides — would be recognized by the human immune system across an entire family of related coronaviruses. The algorithm then assembled those fragments into a single candidate optimized to provoke broad protection.
"This is the first time a vaccine whose active ingredient was designed entirely by artificial intelligence — not just tweaked or analyzed by it — has been tested in humans," researchers said in a statement accompanying the trial. The team chose a nasal delivery route to train immunity at the airways, where coronaviruses typically establish infection.
Why "universal" matters
Most coronavirus vaccines, including the COVID-19 shots rolled out from 2020 onward, are tuned to one virus and need to be updated as variants emerge. A pan-coronavirus vaccine would in theory cover SARS, MERS, COVID-19 and future spillover strains in a single shot — a major prize for pandemic preparedness.
The Phase 1 results don't prove pEVAC-PS prevents infection; safety trials are designed to test tolerability and immune signals, not real-world protection. But the breadth of the response — reacting to multiple coronavirus species at once — is what makes researchers optimistic. Larger Phase 2 trials, which test efficacy in more diverse populations, are the natural next step.
A turning point for AI in medicine
The pEVAC-PS milestone joins a growing list of clinical successes for AI-designed biology. AlphaFold has already reshaped protein research, and computationally designed antibodies and enzymes have moved into trials. But a fully AI-designed vaccine reaching the clinic, and clearing its first safety hurdle, is a new mark.
For public-health officials, the appeal is speed. If AI can shorten the years it normally takes to identify and validate vaccine targets, the world could respond faster the next time a novel coronavirus emerges. As one of the trial authors put it: the goal is to have the next-generation vaccine ready before the next pandemic, not after.
