Radical Numerics AI-designed phage therapy progress

The US-based startup Radical Numerics has reported the administration of an AI-designed bacteriophage cocktail to a human patient suffering from a multi-drug resistant bacterial infection. The treatment, utilizing the company’s Omnii-1 AI model, was conducted under a physician-directed compassionate-use arrangement, reportedly authorized via the FDA expanded access pathway. The viruses were generated to target specific bacterial strains by designing tailored phages based on the genetic characteristics of the target bacteria. This approach aims to bypass the traditional, laborious process of searching environmental collections for matching phages. If confirmed through pending peer-reviewed scientific analysis, the development could offer a scalable solution to antimicrobial resistance, which is projected to cause nearly 40 million deaths by 2050. While the company describes the intervention as a medical milestone, the procedure was not part of a formal clinical trial and lacks a registered trial number, defined research phases, or oversight from an ethics committee. Certain details remain unverified; Radical Numerics has not released clinical outcomes or specific phage sequences, and the FDA has not publicly confirmed the intervention. Additionally, questions have been raised regarding the treatment location, as a collaborator mentioned in reports is based at the Ottawa Hospital in Canada. Following the initial intervention, Radical Numerics reported the completion of what it describes as the world’s first clinical trial using an AI-designed bacteriophage. This procedure targeted a persistent infection caused by an Escherichia coli (E. coli) strain resistant to all traditional antibiotics and existing phage therapies. Using the Omnii computational model, researchers genetically modified a T7 phage to create a custom virus designed to adhere to and destroy the specific bacteria. While the FDA authorized the trial, the company has not disclosed the patient’s identity or specific clinical outcomes due to privacy.