From Code to Cure: AI Generates Successful Bacteria-Killing Virus

By Esha Desai,

The Lawrenceville School, NJ

Imagine a world where people can have access to novel treatment options that are accessible to everyone. Ailments such as bacterial infections like black rot will become a thing of the past. The first steps seen in scientific history towards such a future are illustrated in Stanford University’s breakthrough research discussing the novel bacteriophages created through genome language models in September of 2025. Scientists at Stanford University and the nonprofit Arc Institute collaborated to create 16 AI-written DNA strands that produce working variations of a bacteriophage called phi174. However, it is important to note that this discovery is not an AI-designed life, as viruses are not alive. The AI regenerates genetic code for simple genomes. In this work, the scientist wanted to develop variants of phiX174, a virus that infects bacteria. phiX174 is relatively “simple” as it only has 11 genes and about 5,000 nucleotides. 

The scientists used two versions of an AI called EVO to design phiX174. These AI models are similar to ChatGPT in the way that they train the model on large data sets of genomes of about 2 million other bacteriophages, rather than internet articles. These language models were created using data from genomes of simple life forms such as bacteria and archaea, but also about 150,000 plants or animals. Using this information, the AI printed out possible genomes that are all variants of the phiX174 bacteriophage. To figure out if the genomes worked, the researcher chemically printed 302 of the genome designs as DNA strands into the E coli bacteria. If the genome worked, the E coli would die in the petri dish. Scientists saw a similar outcome of dead bacteria in the petri dish, visible through microscope pictures of the particles, which looked like “fuzzy dots.” In conclusion, 16 viable, novel, and functional bacteriophage genomes out of the 302 designs worked and successfully killed the E coli bacteria. 

In 2008, a lab tried this same experiment, but manually created and printed the genome after reading through scientific literature. Reading literature often takes prolonged periods of time, making AI in biology more attractive to investors. AI will advance science by creating new breakthroughs since it is best at taking information and creating connections between thousands of sources. The creation of a novel genome paves the way for treatment options in humans who have bacterial infections. Gene therapy, the technique of modifying a person's genes to cure a disease, uses viruses to carry into a patient's body. AI may be able to develop viruses that are effective in transferring those genes into the body. 

However, with these future possibilities, it is important to be cautious about viral enhancements, especially to human pathogens, as the consequences for the DNA getting out of control will be severe. Additionally, there are limitations currently to what genomes AI, such as Evo, can produce. For example, AI may not be able to create a genome for more complicated viruses, such as E. coli, which has a thousand times more DNA than phiX174.

The possibility of creating DNA for a bacterium or even a human will be even harder, as scientists will need to change an existing cell, which is labor-intensive. AI will need a larger data set to create complex sequences, and sourcing the right data sets for each project is time-consuming. The success of AI in generating novel genomes for the bacteriophage phiX174 brings further interest into the biotech market, contributing to its huge projection in the future years. Lila Sciences, a biotech company, was able to raise $253 million dollars to build labs run by automated intelligence. AI creating a novel genome of bacteriophages is a new discovery that is changing how biology integrates AI. From code to cure, AI opens a future of creating efficient and novel treatments that transform the biotech industry. 

Works Cited

King, S. H., Driscoll, C. L., Li, D. B., Guo, D., Merchant, A. T., Garyk Brixi, Wilkinson, M. E., & Hie, B. L. (2025). Generative design of novel bacteriophages with genome language models. BioRxiv (Cold Spring Harbor Laboratory). BioRxiv. https://doi.org/10.1101/2025.09.12.675911

Myers, A. (2019). Generative AI Tool Marks a Milestone in Biology | Stanford HAI. Stanford.edu. https://hai.stanford.edu/news/generative-ai-tool-marks-a-milestone-in-biology

Regalado, A. (2025). AI-designed viruses are here and already killing bacteria. MIT Technology Review. https://doi.org/10.1101/2025.09.12.675911v1