ZB ZB
Sport
100 Years of ZB
Live now
Start time
Playing for
End time
Listen live
Listen to NAME OF STATION
Up next
Listen live on
ZB

Listen: 100 Years of ZB Podcast hosted by Mike Hosking

AI used to create viruses not found in nature

Author
Newstalk ZB,
Publish Date
Sun, 23 Aug 2026, 11:54am

The headline sounds like something from a science fiction movie, but the reality is both less alarming and far more scientifically interesting.

Published in the journal Science, researchers designed viruses called bacteriophages, which infect only bacteria with the hope to use them as living medicines to fight infections that no longer respond to antibiotics. 

Scientists have used artificial intelligence to predict proteins and analyse DNA before, but designing an entire working genome with every gene, every control sequence and every interaction needed to build a functioning virus is vastly more difficult. 

The researchers started by training an AI model called Evo on millions of DNA sequences from a tiny bacteriophage called ΦX174, one of the best-studied viruses in biology, which infects only Escherichia coli (E. coli) bacteria. 

The goal was to teach the AI the rules of building similar viruses and it worked - the AI generated around 700,000 completely new viral genomes. 

From these, the scientists selected 285 promising genomes and the DNA was chemically synthesised in the laboratory. 

The synthetic genomes were then inserted into E. coli bacteria and 16 of them successfully produced functional viruses which contained new combinations of genes and DNA sequences that hadn't previously been seen in nature, yet they still managed to infect bacteria successfully. 

The team then tested if these AI-designed phages could solve bacterial resistance. 

They first allowed E. coli bacteria to evolve resistance against the natural ΦX174 virus. 

Normally, once resistance develops, treatment becomes ineffective. 

Instead of using one virus, the researchers mixed all 16 AI-generated phages into a cocktail. 

The bacteria that resisted the natural virus were rapidly overcome by the AI-designed mixture, while a comparable cocktail made from naturally occurring relatives of ΦX174 performed much less effectively. 

Each virus attacked the bacteria in slightly different ways. 

A mutation that protects bacteria from one virus doesn't necessarily protect it from all of them. 

Antibiotic resistance is one of the biggest challenges facing modern medicine and some infections, including tuberculosis and MRSA, are becoming increasingly difficult to treat because bacteria evolve resistance to our existing drugs. 

Phage therapy offers one possible alternative. 

This study suggests that, rather than searching nature for suitable phages, we may one day be able to design entirely new ones on demand, tailored to attack specific bacterial infections. 

That could dramatically speed up the development of future antimicrobial treatments. 

LISTEN ABOVE

Take your Radio, Podcasts and Music with you