Scientists in the United States have for the first time used generative AI to design new viruses not found in nature, according to Anadolu Agency, quoted on Friday, August 7. The study was conducted by researchers from Stanford University and the Arc Institute in California, with results published in the journal Science. The team used a generative AI model trained with millions of natural genomes to design bacteriophages, which are viruses that infect bacteria, in toto. Two genomic language models, Evo 1 and Evo 2, were used, operating similarly to large language models but predicting genetic sequences instead of words. Training data came from genomes of viruses, bacteria, plants, and animals. Researchers chemically synthesized nearly 300 genomes and tested them in the laboratory, producing 16 engineered bacteriophages with different sequences, structures, and survival and growth abilities. When tested, the engineered bacteriophage was more effective in killing E. coli bacteria than the natural virus used as a comparator. The approach could expand synthetic genomics capabilities, and researchers examined its potential for developing phage therapy to fight disease-causing bacteria. They wrote that the approach expands what can be achieved by synthetic genomics, paving the way for adaptive and pathogen-resistant phage therapies. The research remains a preliminary step, with future methods potentially developed for sequencing and genome synthesis in larger biological systems and therapy development.