A team of researchers has used artificial intelligence to design hundreds of new versions of a well-studied virus and then built them from scratch in the laboratory. The work, reported in the journal, demonstrates that AI can generate functional biological agents, a development with both therapeutic promise and potential risks.
The scientists, from Stanford University and the Arc Institute, applied AI models to a phage called ΦX174, which infects a strain of E. coli. They designed new genomes based on this virus and tested them in the lab. Out of 285 AI-designed genomes, 16 produced working phages capable of infecting E. coli.
Out of 285 AI-designed genomes, 16 produced functional phages that could infect E. coli.
The AI systems used in the study are called Evo 1 and Evo 2. The researchers trained these models on a vast amount of genetic data, including more than 2 million phage genomes and nine trillion nucleotides from various organisms such as animals, plants, and microbes. They then fine-tuned the models using genomes from about 15,000 close relatives of ΦX174.
In one notable instance, the AI-generated genome contained a combination of genetic elements that would not have worked in the original phage but worked in the altered genome. The researchers also exposed the AI-generated phages to a variant of E. coli that was resistant to the original phage, resulting in hybrid phages that could infect the previously resistant bacteria.
Designing viruses seemed like a very logical first step.
Medical potential and security concerns
The ability to design viruses with AI could advance medicine, as viruses are often used as delivery vehicles for therapies. Oliver Crook, an organic chemist at Oxford University who was not involved in the study, told the New York Times that many medical doctors see viruses as convenient biological transport vehicles.
However, the same technology raises biosecurity worries. Moritz Hanke, a scientist at the Johns Hopkins Center for Health Security, stated that the model gives criminals the opportunity to construct much more dangerous pathogens. He also noted that the rapid development of synthetic biology has left national regulations far behind.
Brian Hie, a computational biologist at Stanford University, told the New York Times that the research team wanted to be extremely careful in preventing potential risks by deliberately removing all human-infecting viruses from the training data fed to the machine.
