Researchers from Stanford University and the Arc Institute obtained 16 functional bacteriophages from genomes designed with artificial intelligence. The information was published by WIRED this Friday (7), after the study came out in the journal Science.
Bacteriophages are viruses that infect bacteria. In the experiment, the team worked with variants capable of attacking strains of Escherichia coli, with no ability to infect human cells.
Models generated thousands of genomes
The scientists used the Evo 1 and Evo 2 models, developed to analyze and produce genetic sequences. The systems learned patterns related to gene organization, regulatory elements, and structures necessary for a genome to function.
The team used the bacteriophage ΦX174 as a reference, which has 5,386 nucleotides and 11 genes. The virus infects the bacterium E. coli and has a widely studied genetic structure.
The goal was not to copy ΦX174. The models were supposed to create genomes with the necessary organization to recognize the bacterium, insert DNA, replicate the genetic material, and form new viral particles.
After generating thousands of sequences, the researchers selected 285 designs for synthesis and laboratory evaluation. Of that group, 16 produced bacteriophages capable of propagating.
The functional genomes had between 67 and 392 mutations compared with the closest natural viruses. In 13 of them, the team identified changes that did not appear in the natural sequences consulted.
The 16 viruses infected the non-pathogenic strains E. coli C and E. coli W. The tests did not record propagation in six other bacterial strains, according to the Arc Institute's technical report.
Viruses circumvented bacterial resistance
The scientists also evaluated the bacteriophages against three strains of E. coli that had developed resistance to ΦX174. Cocktails made with the AI-designed viruses established new infections after one to five passages.
ΦX174 used alone could not overcome the resistance. The researchers attributed the result to the genetic diversity of the artificial bacteriophages, which offered different paths to attack the bacteria.
The data point to a possible future application in phage therapy, an approach that uses viruses against bacterial infections. The work, however, remains in the experimental phase and did not involve animals or patients.
Research expands debate on biosecurity
The experiments used non-pathogenic bacteria and containment protocols. According to the Arc Institute, the training data also excluded sequences of human viruses.
Even so, the ability to design complete viral genomes expands the debate over the misuse of these tools. Researchers in the biosecurity field argue that control mechanisms should keep pace with the evolution of models and genetic synthesis techniques.
The results presented do not demonstrate the creation of viruses capable of infecting people. The study serves as a proof of concept for the development of bacteriophages and for research aimed at antibiotic-resistant bacteria.
