Radical Numerics Administers First AI-Designed Phage to Human Patient

Key points

• Radical Numerics used its Omnii-1 model to edit the tail fiber gene of the T7 phage, creating variants that target specific bacterial receptors.
• The treatment was given to a patient with an E. coli infection resistant to both antibiotics and existing phage therapies.
• The company reported that 69.9% of designed variants were viable, compared to 4.8% for random controls, and 2,161 variants showed improved activity.
• Clinical details, including the patient’s outcome and hospital, have been withheld to protect privacy and prevent identification.
• The FDA authorized the administration under a compassionate-use pathway, though the therapy remains unapproved for broader clinical use.

Background

Bacteriophage therapy has been used for over a century but faces challenges in matching specific phages to bacterial strains. Previous AI efforts, such as those by the Arc Institute, generated viable phages in labs but did not involve human patients. The current development follows a 2019 case where engineered phages treated a drug-resistant Mycobacterium abscessus infection, but this is the first time AI has designed the therapeutic agent. The broader context includes rising antimicrobial resistance, which caused approximately 4.7 million deaths in 2021, and concerns about AI-enabled biological design tools.

How outlets are covering it

TechRadar and The Independent emphasize the novelty of AI-designed phages entering human medicine, highlighting the potential to speed up treatment for resistant infections. Contagion Live provides technical details on Omnii-1’s performance, noting it outperformed other prediction systems in identifying active phages. FOX 17 West Michigan News offers a broader view of AI in medicine, linking this development to other AI health tools like cancer survival prediction. Experts like Harvard microbiologist Michael Baym express skepticism until peer-reviewed data is available, while Radical Numerics CEO Eric Nguyen commits to future publication. The company’s decision to withhold clinical details has drawn criticism from some outlets but is defended by the company as necessary for patient privacy.

Why it matters

This development could accelerate the treatment of antibiotic-resistant infections, which are projected to cause nearly 40 million deaths between 2025 and 2050. AI-driven phage design may reduce the time required to find effective treatments from months to days. However, the lack of peer-reviewed data and transparency raises concerns about safety and efficacy. The dual-use nature of AI in biological design also prompts calls for stronger biosafety measures.

What to watch

Radical Numerics plans to publish its findings through peer review, though the timeline is unclear. The company intends to commercialize the proprietary treatment. Regulatory bodies and the scientific community will likely scrutinize the case as more details emerge. Broader discussions on AI biosafety and the governance of biological design tools are expected to intensify as AI capabilities in this field advance.