COULD AI 'SOLVE' VIRUSES BEFORE MATHS?

Much has been written about AI ‘solving maths’. Even ‘solving physics’. But might it help solve viruses first?

A Cambridge-led team has taken a new universal vaccine technology, developed with AI, into human testing for the first time.

Current vaccine cycles start with strains already circulating. Jonathan Heeney, who led the work, says it’s like ‘a dog chasing its tail'.

His team tried the reverse. They took genetic data from across the sarbecovirus family – SARS, Covid and related viruses circulating in bats – and used digital intelligence to find the features that stay relatively stable as those viruses mutate. And from those shared features, they designed a synthetic 'super-antigen'.

The aim is a single vaccine that can train the immune system against many variants at once, including viruses that haven't yet crossed into humans.

Of course the slower step is validation.

This antigen was designed in 2020. Six years on, a first safety trial is as far as it’s come.

The first human trial cleared the safety bar: 39 volunteers received the vaccine, and it was safe and well tolerated across all four dose levels.

It also gave a positive signal. Antibodies from a small group of participants homed in on the conserved, cross-family region the antigen was built to target across sarbecoviruses.

The overall immune response was modest and variable, and the trial ran through the Omicron years, when participants arrived with different levels of existing Covid immunity. Larger studies will need to show whether the response is strong, durable and broad enough to protect people.

Even so, it's a real reason to be optimistic. Digital intelligence designed a vaccine around the shared biology of a viral family, rather than one known strain – and the same platform is now aimed at flu and the Ebola group.