Artificial intelligence has been used to design complete viral genomes capable of producing functional viruses that can replicate under laboratory conditions, marking a significant development in the rapidly growing field of AI-driven biology.
Researchers in the United States used generative AI models to create entirely new bacteriophages — viruses that specifically infect bacteria. The resulting viruses were designed to target bacteria rather than humans and were studied under controlled laboratory conditions.
The research represents an important step beyond previous uses of artificial intelligence in areas such as discovering antibiotics, predicting protein structures and identifying potential drug candidates. Researchers say it demonstrates that generative AI can potentially design much more complex biological systems.
The technology works in a way that has similarities to large language models. While systems such as ChatGPT learn patterns in language and predict sequences of words, biological AI models can learn patterns contained within genetic sequences.
Researchers used AI models known as Evo 1 and Evo 2, which were trained using large amounts of genetic information. The systems were then adapted for research involving bacteriophages.
Scientists generated hundreds of potential designs before selecting candidates for laboratory testing. According to the research, 16 of the synthesised designs successfully demonstrated the ability to infect and kill E. coli bacteria.
The results suggest artificial intelligence could eventually become an important tool in developing new approaches to bacterial infections, particularly as antimicrobial resistance makes some existing antibiotics less effective.
Bacteriophages have long attracted scientific interest because they naturally attack bacteria. Researchers are exploring whether carefully selected or engineered phages could provide alternative treatments for certain bacterial infections.
The implications of the research, however, extend beyond bacteriophages.
AI’s growing ability to analyse and generate biological sequences could accelerate synthetic biology, an area of science focused on designing or modifying biological systems for useful purposes.
Potential applications could eventually include developing new medicines, designing therapeutic proteins, creating specialised enzymes and accelerating research into treatments for genetic diseases.
The development also raises important questions about safety.
As AI becomes increasingly capable of designing biological material, researchers and governments will need to consider how such technology should be controlled and how potentially dangerous applications can be prevented.
Scientists involved in this type of research have emphasised the importance of safeguards. In this study, the work focused on bacteriophages rather than viruses that infect humans, and potentially higher-risk viral material was excluded from parts of the research process.
Experts have argued that advances in generative biology should be accompanied by stronger biosafety and biosecurity standards to ensure increasingly powerful AI systems are used responsibly.
There are also major technical limitations.
Designing a relatively small viral genome is very different from creating a living organism. Viral genomes can be dramatically smaller and less complex than the genomes found in living cells, while the human genome contains billions of DNA base pairs.
Nevertheless, the research represents another example of artificial intelligence moving beyond generating text, images and computer code into the physical sciences.
AI systems are increasingly being applied to chemistry, medicine, genetics and biotechnology, where their ability to analyse enormous datasets could significantly accelerate scientific discovery.
The development could ultimately prove valuable for healthcare and biotechnology, particularly if researchers can use AI to design biological tools capable of addressing diseases that are difficult to treat using existing methods.
At the same time, the ability to computationally design increasingly sophisticated biological systems means safety will have to develop alongside capability.
The emerging era of generative biology could therefore become one of artificial intelligence’s most important — and closely scrutinised — scientific frontiers.
