Researchers have reported the first human trial of a vaccine whose active component was designed entirely by computer simulations. Published in the June 2026 issue of the Journal of Infection, the phase I study tested pEVAC-PS — a DNA vaccine built around an AI-designed “super-antigen” intended to elicit immunity across a broad family of coronaviruses related to SARS and COVID-19.
What the trial tested
The trial ran in the U.K. from December 2021 through September 2023 and enrolled 39 healthy adults aged 18–50. All participants had previously received two or three doses of COVID-19 vaccine but had no recent SARS-CoV-2 infection. The volunteers were split into four dose groups (0.2 mg, 0.4 mg, 0.8 mg, and 1.2 mg) and received two intradermal administrations: day 0 and day 28.
Instead of a needle injection, pEVAC-PS was delivered with a needle-free intradermal device (PharmaJet Tropis). That delivery method is intended to simplify administration and storage, which could be an advantage in low-resource settings.
Safety and side effects
pEVAC-PS was well tolerated. No serious adverse events were reported. Most reactions were mild or moderate and typical of vaccination — local soreness or redness and transient fatigue or malaise. Higher doses did not produce more side effects, and side effects were generally fewer after the second dose. A small number of participants contracted mild, unrelated COVID-19 during the study and required no medical care.
Immune responses observed
Overall immune response signals were modest. Because participants had varied prior vaccination and infection histories, it was challenging to isolate the vaccine’s direct effect on antibody levels. The highest dose group (1.2 mg) showed a small but statistically significant rise in antibodies targeting the vaccine’s designed spike region about six weeks after the first dose.
Neutralizing activity increased in some middle- and high-dose participants against certain SARS-CoV-2 variants — notably Delta and Omicron BA.1 — but not against the original Wuhan strain or SARS-CoV-1. A detailed peptide microarray showed that participants developed antibodies that recognized conserved regions within the receptor-binding domain (RBD), including the epitope targeted by the broadly neutralizing S309 antibody. While recognition of these conserved sites is encouraging, it did not translate into broad, strong neutralization at the doses tested.
What this means
The trial demonstrates that an AI-designed antigen can be manufactured, delivered without needles, and safely given to humans. It also shows that such a design can direct immune recognition toward conserved RBD regions shared across coronaviruses. However, at the doses used in this small phase I study, the vaccine generated only modest increases in antibody levels and limited neutralizing breadth. Larger trials and further optimization will be needed to determine whether pEVAC-PS (or similar AI-designed constructs) can provide robust, broad protection across diverse populations.
Why AI was used and future implications
AI was applied to search an enormous space of possible antigen sequences and identify candidate regions likely to induce broadly reactive immunity — a task that would be infeasible by brute-force laboratory screening. Proponents say AI can greatly accelerate vaccine design, reduce the time from concept to human testing, and tackle hyper-mutating pathogens by prioritizing conserved, functionally important targets.
Experts quoted in connection with the study note that this approach could be a reusable blueprint for other mutable viruses, potentially aiding universal influenza vaccines, new HIV strategies, or early responses to future zoonotic threats. But they also emphasize that safety, efficacy, and durability must be proven in larger, more diverse trials before clinical use.
Bottom line
pEVAC-PS is the first human-tested vaccine whose active component was fully designed with AI. The phase I trial shows it is safe and can steer immune recognition toward conserved coronavirus structures, but immune responses were modest and neutralizing breadth limited at the tested doses. Further research, dose optimization, and larger trials are needed to assess whether AI-designed vaccines can deliver the broad, durable protection against current and future coronaviruses that researchers hope for.
