Representation of Evo-?36, one of the bacteriophages created with artificial intelligence and tested in the laboratory by researchers. Arc Institute/Samuel H. King et al. For the first time, scientists managed to get artificial intelligence to design the complete genome of a virus — not just an isolated gene, but the entire set of genetic instructions needed for the organism to function, reproduce and infect a host. The study, published this Thursday (6) in the prestigious scientific journal "Science", was conducted by researchers from the Arc Institute and Stanford University, in the United States, and is considered a milestone in so-called synthetic biology. The team used an AI system called Evo, which works in a similar way to ChatGPT and other language models: instead of learning to predict the next word in a sentence, Evo learned to predict the sequence of DNA "letters" — the bases A, C, G and T that make up our genetic code. ? To do this, he was fed the genetic material of millions of bacteria, plants, animals and viruses. ?Download the g1 app to see news in real time and for free After training, the system was directed to create bacteriophage genomes, a type of virus that exclusively infects bacteria and is harmless to humans. As a model, the scientists used ?X174, a small and extremely well-studied virus, whose genome has around 5,400 letters, much less than the approximately 500,000 letters in the genome of the simplest bacteria, and a universe away from the 3 billion letters of human DNA. ? UNDERSTAND: DNA works like a large instruction manual: what changes from one organism to another are not the “letters” available, but the order in which they appear. The genome is the complete set of these instructions, including the sections that guide when and how each part should be used. AI has generated hundreds of thousands of genome proposals. The researchers selected the most promising ones, ordered the synthesis of almost 300 of them in the laboratory and tested them against E. coli bacteria. The result: 16 of these computer-created viruses proved to be viable, that is, they were able to multiply and destroy bacteria, just as a natural virus would do. READ ALSO: Why scientists believe in the existence of oceans on dwarf planets Book bound with human skin is displayed at literary fair in NY How whales can teach scientists to talk to aliens
For the first time, AI creates viruses that did not exist in nature; understand the advancement and risks
Representation of Evo-?36, one of the bacteriophages created with artificial intelligence and tested in the laboratory by researchers. Arc Institute/Samuel H. King et al. For the first time, scientists managed to get...
It's as if the AI ??has learned the "grammar" of life — the rules that make one gene work correctly alongside another — without ever being explicitly taught them, just by observing millions of ready-made examples. The most striking of the results, however, was another: when scientists exposed the artificial viruses to bacteria that had developed resistance to the natural virus ?X174, a combination of the AI-generated viruses managed to overcome this resistance — something that a cocktail of similar natural viruses was unable to do. This suggests that, in the future, the technology could help create treatments against bacterial infections that no longer respond to antibiotics, a growing problem in medicine. ? "Groups undertaking this type of complete genome design work should consult biosafety professionals throughout the project," said the authors in the study. This caveat is not surprising. At the same time as they celebrate the progress, the researchers themselves and other experts point out risks. If an AI can reconstruct the rules that allow it to assemble a functional virus from scratch, there is concern that, in the future, similar technology could be used to design more dangerous pathogens, capable of infecting people, animals or plants. Therefore, the team adopted a series of precautions. During training, the system did not receive any genetic information from viruses that infect humans, and sequences from viruses that attack other animals, plants or fungi were also excluded on purpose. ?? All work was carried out at the laboratory safety level recommended for bacteriophages, considered low risk. Image compares the internal structure of Evo-?36, created with artificial intelligence, with that of a virus found in nature. In purple, one of the proteins used to store DNA appears. Samuel H. King et al. Still, scientists linked to the area of ??biosafety say that governance over this type of technology — laws, protocols and ways to monitor access to AI tools and DNA synthesis — has not yet advanced at the same pace as science. “This is impressive work as a proof of concept,” Simon Jackson, a researcher at the University of Waikato who was not involved in the study, told Science Media Centre. “Designing larger bacteriophages, reliably controlling which bacteria they infect, and proving safety and efficacy in patients are steps that still lie ahead,” he added. It is worth remembering that the viruses created in this experiment are far from representing a threat to human health: they can only infect a specific and harmless strain of E. coli used in the laboratory. It is also not about "artificial life", viruses are not even considered living beings by most scientists, as they cannot reproduce on their own, only within a host's cells. To reach truly living organisms, even the simplest ones, researchers recognize that a much greater leap in complexity would still be necessary. VIDEO: Why is the James Webb a super telescope? Compare photos of the James Webb super telescope with its predecessor
This story was originally published by G1. Visit the original publication for further details.
Open original publication
