AI-designed viruses are here and already killing bacteria

Artificial intelligence can draw cat pictures and write e -mails. Now the same technology can put together a functioning genome. A research team in California said that KI used to propose new genetic codes for viruses - and it was possible to bring several of these viruses to replicate and kill bacteria. The scientists based at Stanford University and the non-profit ARC Institute, both in Palo Alto. The work described in a form paper has the potential to create new treatments and to accelerate research into artificially constructed cells. It is also an "impressive first step" in the direction of AI-Entworten life forms, says Jef Boeke, biologist at the Nyu Langone Health, which was made available for a preliminary copy of the paper by with Technology Review. Boeke says that the performance of the AI ​​was surprisingly good and that his ideas were unexpected. "They saw viruses with new genes, shortened genes and even different regulations and arrangements," he says. However, this is not yet a life of AI. This is because viruses are not alive. They are rather renegade bits of the genetic code with relatively measly, simple genomes. In the new work, researchers from the ARC institute, variants of a bacteriophage - tried to develop a virus that infected bacteria - Phix174, which only has 11 genes and about 5,000 DNA letters. For this purpose, they used two versions of a AI called Evo, which works according to the same principles as large language models such as chatt. Instead of feeding textbooks and blog posts to learn them, the scientists trained the models about the genomes of around 2 million other bacteriophage viruses. But would the genomes proposed by the AI ​​make sense? In order to find out, the Californian researchers chemically printed 302 of the genome designs as DNA strands and then mixed those with E. coli bacteria. This led to a profound moment "Ki is here", in which the scientists saw plaque dead bacteria in their petri dishes one night. Later they made microscope pictures of the tiny viral particles that look like blurred points. "It was pretty striking to see this sphere just like this," says Brian Hie, who lists the laboratory of the ARC Institute in which the work was carried out. A total of 16 of the 302 designs were called that the computer-aided phage began to replicate, finally broke through the bacteria and killed them. J. Craig Venter, who created some of the first organisms with laboratory DNA almost two decades ago, says that the AI ​​methods for him are "only a faster version of test and error experiments". For example, when a team that he headed, in 2008, made a bacterium with a laboratory dimmage, it was after a long hit-or missed process, in which different genes were tested. "We made the manual AI version - equipped with the literature that took what was known," he says. But exactly the speed is why people will change biology. The new methods already claimed a Nobel Prize in 2024 to predict protein forms. And investors are billions that AI can find for new medication. This week, a company collected 235 million US dollars for the construction of automated laboratories that are operated by artificial intelligence. Computer -aided viruses could also find commercial purposes. For example, doctors have sometimes tried to treat patients with serious bacterial infections. Similar tests are in progress to cure cabbage made of black rot, which are also caused by bacteria. "There is definitely great potential for this technology," says Samuel King, the student who led the project in Heis's laboratory. He notes that most gene therapies use viruses to bring genes into the patient's body, and AI could develop more effective. The Stanford researchers say they have deliberately not taught their AI about viruses that people can infect. However, this type of technology creates the risk that other scientists - curiosity, good intentions or malice - can turn the methods of human pathogens and examine new dimensions of lethality. "An area in which I ask the extreme caution is a research for viral improvements, especially if it is accidental so that you don't know what you get," says Venter. "If someone had done this with smallpox or anthrax, I would have serious concerns." An open question remains whether a AI can create a nasty genome for a larger organism. For example, E. Coli has about a thousand times more DNA code than Phix174. "The complexity would be breathtaking from breathtaking to ... much more than the number of subatomar particles in the universe," says Boeke. In addition, there is still no easy way to test AI designs for larger genomes. While some viruses can only "start" from a DNA strand, this is not the case with a bacterium, a mammoth or a person. Instead, scientists would have to change an existing cell with genetic engineering - an even tedious process. Nevertheless, Jason Kelly, CEO of Ginkgo Biowork, a cell carrier in Boston, says exactly such an effort. He believes that "automated" laboratories could be carried out in which genomes are proposed and tested and the results are attributed to AI for further improvement. "This would be a scientific milestone in the national in advance, since the cells are the building blocks of all lives," says Kelly. "The United States should make sure that we will get there first."