Researchers at the University of Cambridge and its spin-out DIOSynVax have completed the first human trial of pEVAC-PS, a coronavirus vaccine whose active component was designed entirely by AI. The Phase 1 study, run with 39 healthy volunteers at NIHR facilities in Southampton and Cambridge, found the shot safe with no significant side effects. Results were published this week in the Journal of Infection.
What the AI actually did
The pitch here is genuinely different from how vaccines normally get made. Instead of grabbing an antigen from a virus strain already circulating, the team fed machine learning every Sarbeco coronavirus genetic sequence logged by global surveillance programs, then had it design a synthetic super-antigen built from features the whole family shares. SARS-CoV-2, the original SARS virus, plus a batch of bat coronaviruses that have not jumped to humans yet.
The Cambridge announcement calls this the first vaccine whose active component was designed entirely by computer simulation to reach a human trial. The trial paper lays out the dosing: four escalating cohorts at 0.2, 0.4, 0.8, and 1.2 mg, given on day zero and day 28, delivered as a needle-free DNA shot through a jet injector.
The number nobody is putting in the headline
Safety is what a Phase 1 is for, and pEVAC-PS cleared that bar. The immune response is where things get murkier. Volunteers had all received two or three doses of existing COVID vaccines, so they walked in with substantial immunity already. Against that backdrop the response to the new shot was, in the trial's own framing, modest. Measurable, but modest.
That matters because the whole selling point is breadth. The vaccine did stimulate responses to SARS and to bat viruses that have never infected a person, which is the interesting part and hard to fake. Whether that translates into actual protection is a question this trial was never built to answer.
"We've converted vaccine development from being reactive to being future proof," said Cambridge's Jonathan Heeney, the scientific lead. It is the kind of line that sounds great until you remember a 39-person safety study cannot prove future-proofing of anything.
So what happens next
A larger Phase 2 trial will test whether the vaccine produces strong, broadly protective responses across a more diverse group, which is the real test of the platform. DIOSynVax, founded in 2017 out of Cambridge Enterprise, is also pointing the same AI approach at flu and hemorrhagic fever viruses including Ebola.
The work was funded primarily by Innovate UK. No date has been set publicly for the Phase 2 start, so the next concrete signal will be enrollment details for that trial.
