AI-Designed Pan-Sarbecovirus Vaccine Clears Phase I Safety Bar, Stirs Broad Coronavirus Antibodies

In a first-of-its-kind human trial, the computationally built pEVAC-PS shot was well tolerated at all four dose levels and prompted antibodies that recognize conserved regions shared across SARS-CoV-2 variants, SARS-CoV-1, and bat relatives — though efficacy in humans remains unproven.

A vaccine antigen designed entirely by computer has completed its first-in-human Phase I test, demonstrating a clean safety profile and early signs that it can coax the immune system toward the broad cross-coronavirus recognition its designers intended.

The candidate, pEVAC-PS, was built by the University of Cambridge and its spin-out DIOSynVax using a computational platform that maps structurally conserved epitopes across the sarbecovirus subgenus — the coronavirus lineage that includes SARS-CoV-2, SARS-CoV-1, and a wide range of bat coronaviruses with zoonotic spillover potential. Rather than target a single strain’s spike sequence, the approach aims for shared vulnerabilities that variants are unlikely to mutate away.

The open-label, dose-escalation trial enrolled 39 healthy adults aged 18–50 at two NIHR facilities in Southampton and Cambridge between December 2021 and September 2023. Participants received two doses of pEVAC-PS delivered intradermally via the PharmaJet Tropis needle-free jet injector across four escalating dose cohorts. The primary endpoint was safety and tolerability; immunogenicity was secondary.

The vaccine was well tolerated at all four dose levels, with no significant safety concerns reported — a critical hurdle for any novel vaccine platform entering humans for the first time.

On immunogenicity, the highest-dose cohorts generated neutralizing antibody responses against SARS-CoV-2 Delta and Omicron variants and showed increased reactivity at RBD regions — including a conserved epitope corresponding to that bound by the broadly neutralizing monoclonal antibody S309 — consistent with the vaccine’s cross-sarbecovirus design goal. The authors note that interpretation was complicated by high pre-existing antibody levels and variable Omicron exposure histories during the trial period.

The researchers describe pEVAC-PS as the first AI- and computer-simulation-designed vaccine to complete human Phase I testing.

Important caveats apply. This was a small Phase I study powered to assess safety and preliminary immunogenicity only. No efficacy data — whether the vaccine prevents infection or disease — exist in humans. The trial is published in the Journal of Infection (2026).

Trials Today

Phase III SKYSCRAPER-15: Tiragolumab + atezolizumab vs. placebo + atezolizumab in resected PD-L1+ NSCLC (n=56) — Terminated — primary efficacy objective abandoned, converted to safety-only follow-up. Fourth consecutive TIGIT Phase 3 failure for Roche.

Phase III INTEGRATE IIb: Regorafenib + nivolumab vs. chemotherapy in refractory gastro-oesophageal adenocarcinoma (n=462) — Results first-posted to ClinicalTrials.gov this week. International trial; OS data now available.

Phase III DanNORMS: Rituximab vs. ocrelizumab in active MS (n=600) — Primary completion June 10, 2026 — data lock imminent. Non-inferiority result would challenge $65,000/year ocrelizumab pricing.

Phase III REIMAGINE 5: CagriSema vs. tirzepatide 5mg in type 2 diabetes (n=1,023) — Primary completion June 2, 2026. First head-to-head of two leading obesity/T2D platforms; topline results awaited.