Scientists Train AI on DNA and It Creates 16 New Viruses

Scientists at Stanford University have used an artificial intelligence model trained on millions of DNA sequences to design new viral genomes.

According to Forbes, 16 of the AI-generated viruses successfully infected E. coli in laboratory experiments.

The findings demonstrate that an AI model can learn patterns and evolutionary constraints from large collections of genetic material and use that knowledge to generate functional viral genomes that do not match natural sequences.

The researchers trained an OpenAI model called Evo on extensive genomic data, similar to how language models learn patterns from large collections of text. The model was then used to generate new genetic sequences that scientists synthesized and tested in the laboratory.

Of the AI-designed genomes that were synthesized and tested, 16 produced functional bacteriophages capable of infecting E. coli. Some were also able to overcome natural resistance mechanisms used by the bacteria.

The researchers said they excluded human pathogen data from the training process, meaning the viruses produced in the study were not designed to infect humans.

Biosecurity experts have warned that the ability to generate functional biological sequences could eventually lower barriers to designing harmful pathogens if adequate safeguards are not introduced.

Get the latest tech news, telecom insights, and product launches wherever you prefer.