The Zoonotic Viruses Laboratory in the University of Cambridge's Department of Veterinary Medicine published a paper in Cell Reports Medicine on November 5 describing an H5 avian influenza vaccine candidate designed with the help of AI. In mouse and ferret experiments, it gave complete protection against the range of variant strains tested. As a comparison, the vaccine stockpiled by the World Health Organization, facing the same viruses, left only about half of the animals alive.
The study was led by Professor Jonathan Heeney, with Dr Sneha Vishwanath involved, and partners include the Cambridge spin-out DIOSynVax, Canada's National Microbiology Laboratory and Italy's IZSVe institute. DIOSynVax was founded in 2017, and its name is short for Digitally Immune Optimised Synthetic Vaccines.
Which step the AI handled
The team built a tool that pools gene sequences from avian flu surveillance around the world, drawn from wild birds, livestock and human cases, and then used it to design a set of "super-antigens" carrying the features this family of viruses has in common. The candidate antigens were first screened in cells, and the best performer moved on to animal experiments. Cambridge's summary does not name the model that was used.
Heeney described these antigens this way:
"The broadly protective antigens we design are entirely novel – they don't exist in nature."
Vishwanath explained that the vaccine trains the immune system to target several different stages of the virus's life cycle at once. Heeney also noted that AI greatly sped up antigen design, but each candidate it produces still has to be safety-tested in the lab one by one.
Compared with stockpile and seasonal vaccines
The WHO's stockpiled vaccines are designed around a single strain. According to Cambridge, they offer only limited protection against other strains now circulating, and the comparison results in this animal study point the same way.
Seasonal flu vaccines take a different route: predict first. Twice a year the WHO recommends strains for the coming season based on surveillance data, and manufacturers produce to that recommendation. If the forecast is off, protection that season is weaker. H5 avian flu is harder to predict still: it moves between birds, cattle and other mammals, with many branches and fast change.
The Cambridge approach takes a different line of thinking, using one antigen to cover a whole group of viruses as far as possible. Heeney described the design goal as offsetting viral evolution, so that researchers need not chase each mutation. The team also offered a supporting data point: the seasonal flu antigen they designed in 2018 showed, in preliminary results, that it was still effective seven years later. These are preliminary data, and the summary gives no specific protection rate.
What remains unanswered
All the evidence so far comes from animals. Cambridge's summary does not give the number of animals per group, the statistical methods, or which variant strains were tested; those will have to be checked against the full paper. The next step is human clinical trials, for which no timetable has been announced. Whether the vaccine is safe in people and whether it can produce the same protection there has not yet been tested.
On funding, the team received £4.28 million in October from UK Research and Innovation (UKRI) and the Department for Environment, Food and Rural Affairs (Defra), to develop, together with partners in Regensburg and Southampton, a vaccine against Lassa fever, Marburg virus and Ebola virus at the same time. That money belongs to a separate line of research and is unrelated to this bird flu vaccine.
Heeney said the moment that worries them most is when avian flu adapts to survive in the respiratory tract of mammals. What AI shortened in this study was the antigen design step; how long clinical trials, approval and mass production will each take is not addressed in the paper.
Sources: University of Cambridge official news, Cell Reports Medicine paper information, CocoLoop; verified animal-experiment protection results, the WHO stockpile vaccine comparison, and the UKRI and Defra funding amount.