American startup uses AI-designed viruses to treat diseases

According to the New York Times, startup Radical Numerics described in detail the virus development process in a technical report posted on the company website on October 20. They refused to provide information about the patient as well as the doctor conducting the test, citing confidentiality reasons.

According to Futurism, although this report has not been approved or published in a scientific journal, company leaders said the test has been licensed by the US Food and Drug Administration (FDA). "We plan to publish our scientific findings in specialized journals, but publishing clinical details later requires consent from patients or competent authorities," Eric Nguyen, CEO of Radical Numerics, shared.

If successful, the experiment will open a new direction to apply AI in medicine, helping to devise individual treatment regimens for each patient to destroy even the most stubborn drug-resistant bacteria.

AI-engineered viruses may be new, but phage therapy has been around for more than a century. Chosun Daily said, the virus will attach to the surface of the bacteria and inject its genetic material inside. Here, the genetic code will multiply continuously, causing the bacteria to swell, burst and die. Bacteriophages have attracted a lot of attention as a new alternative treatment to antibiotics to kill drug-resistant superbugs thanks to their ability to target extremely precisely without damaging other healthy cells in the body.

Today, as the problem of antibiotic resistance worsens globally, this treatment method is having a resurgence, but it is much more difficult to carry out than oral medications. Each type of phage can only kill certain strains of bacteria. So when one becomes available, doctors must test whether it is actually effective against the patient's infection. Dr. Marisa Azad, an infectious disease physician at Ottawa Hospital in Canada who is collaborating with Radical Numerics, admits the hunt for bacteriophages is time-consuming.

Some researchers are trying to overcome that barrier by looking for new types of bacteriophages in soil and water. But the advent of AI spurred the idea of ​​designing new bacteriophages that could be customized to fight the specific bacteria causing a patient's infection.

In 2025, the team at the Arc Institute in Palo Alto, California created the Evo 2 AI model and trained it on millions of DNA sequences, including bacteriophage genes. They asked Evo 2 to create new genomes similar to the genome of a group of viruses that specialize in killing bacteria. As a result, AI created 700,000 virus genomes. Researchers synthesized DNA molecules from 285 genomes and inserted them into bacteria. They discovered 16 genomes that create viable viruses.

Three members of the Evo team, including Dr. Nguyen, left to form Radical Numerics. In June, the company launched a new AI model called Omnii, which works on some of the same principles as Evo. Not limited to DNA, Omnii can be trained on other aspects of biology such as protein structures and molecules that help turn gene activity on/off. Radical Numerics is using Omnii to solve a range of challenges, including designing cancer vaccines and identifying the risk of pandemic viruses. Among these efforts is the first trial of an AI-engineered phage therapy.

Researchers created several variations of bacteriophage T7. This virus that only attacks E. coli bacteria is shaped like a miniature Moon lander with a spherical protein shell containing DNA, placed on 6 legs sticking out around it. To attack bacteria, T7 will land on the bacterial surface, use its legs to cling tightly to the position, and then inject its DNA into the bacterial cell. Dr. Nguyen and his colleagues said that Omnii creates new viral genes that change the shape of T7's legs. Although some of the new legs differ significantly from the natural legs, Omnii's phages can still infect bacteria.

Once they were sure the method was successful, the team collaborated with a doctor to treat a patient with a persistent E. coli infection caused by bacteria that were resistant not only to antibiotics but to all available bacteriophages. They trained Omnii with information about resistant strains and asked it to design T7 with legs that could latch onto bacteria. AI provides the genome for the new virus and scientists create the corresponding DNA molecule and insert it into the bacteria. Infected bacteria produce bacteriophage T7 that can effectively attack and destroy other bacteria. After testing to ensure the virus was safe, Radical Numerics introduced it to doctors and the FDA granted permission to conduct testing.

Phage therapy can last days or weeks, and doctors often monitor their patients' condition for months to guard against recurring infections. Therefore, it is necessary to wait a while to determine whether the virus designed by Omnii works well or not.

If the trial is successful, Radical Numerics hopes to contribute to accelerating phage therapy by significantly shortening the time to create new viruses tailored to each patient's infection characteristics.