Cambridge tests AI-designed universal coronavirus vaccine

Researchers at the University of Cambridge and its spinout DIOSynVax (DVX) Ltd have completed the first human trial of a vaccine whose active antigen was designed entirely by computer, according to a University of Cambridge press release, the NIHR, and a peer-reviewed report in the Journal of Infection. The Phase 1, dose-escalation trial of the candidate pan-Sarbecovirus vaccine pEVAC-PS enrolled 39 healthy volunteers aged 18 to 50 at NIHR Clinical Research Facilities in Southampton and Cambridge, and was sponsored by University Hospital Southampton NHS Foundation Trust. The sponsors reported the vaccine was safe and caused no significant side effects.

According to the Cambridge release and ScienceDaily, the team used machine learning to analyse genetic sequences of Sarbeco coronaviruses gathered through global surveillance, then combined the broadly conserved features into a single compact immunogen they call a "super-antigen." In this trial the antigen was delivered as a DNA vaccine through a needle-free micro fluid jet. The work is described as the first time an antigen created entirely through computer simulations has been tested in people.

The trial team reported cross-reactive immune responses not only to SARS-CoV-2 and SARS but also to related bat sarbecoviruses that have not infected humans, per the Cambridge release and ScienceDaily. Independent coverage is more measured: AFP reporting (via Medical Xpress) and the paper itself indicate the responses were modest and did not yet demonstrate broad or robust neutralising activity, with the authors framing the cross-reactive binding as support for the design concept rather than proof of protection. This is a safety and proof-of-concept result, not evidence of efficacy.

For machine-learning teams, the milestone is that a sequence-based generative design progressed end-to-end into a human candidate, not merely into hypothesis generation or epitope mapping. Industry observers have seen comparable computational-first workflows compress candidate discovery; the open question, as here, is whether in silico breadth survives contact with immunogenicity, delivery, and manufacturing constraints.

Look for published immune-phenotyping and neutralisation data in the Journal of Infection report (DOI 10.1016/j.jinf.2026.106759), the planned Phase 2 study in a larger and more diverse cohort, and independent replication of cross-reactive responses against diverse sarbecoviruses. Chief investigator Professor Saul Faust of the University of Southampton described the class as "future-proofed," while lead researcher Professor Jonathan Heeney said, "We've converted vaccine development from being reactive to being future proof."

The result validates the feasibility of pushing AI-designed immunogens into human trials, but broader impact will hinge on robust immunogenicity endpoints, scalable manufacture, and replication in larger controlled studies. The project was funded primarily by Innovate UK.