For the first time, researchers have administered viruses designed by artificial intelligence (AI) to treat a human patient with a severe bacterial infection. The viruses, known as phages, specifically target and kill bacteria, offering a potential alternative to traditional antibiotics amid rising concerns about antibiotic resistance. The treatment was conducted under a trial authorized by the U.S. Food and Drug Administration (FDA).
The biotech start-up Radical Numerics, which created the AI-designed phages, published a technical report on its website outlining the development but withheld details about the patient and the clinical team, citing confidentiality and patient privacy. The company said it plans to publish peer-reviewed research on the therapy in the future. Eric Nguyen, the company’s CEO, emphasized their commitment to protecting patient confidentiality and maintaining clinical trust.
Phage therapy, a concept dating back more than a century, has regained interest recently as antibiotic-resistant infections escalate worldwide. Unlike antibiotics, which can harm beneficial bacteria alongside harmful ones, phages are highly specific, infecting only particular bacterial strains. However, sourcing effective phages has traditionally been time-consuming because each variant targets only limited strains, requiring extensive screening to find a match for a patient’s infection.
To address this challenge, Radical Numerics used an AI model called Omnii to rapidly design phages tailored to the resistant strain infecting the patient. The AI was trained on vast biological data, including DNA sequences and protein structures, enabling it to generate new viral genomes that alter the phage’s bacterial-binding structures. These genomes were synthesized, inserted into bacteria, and produced viable phages capable of infecting and destroying the resistant bacteria.
The AI-designed phages are based on T7, a virus that specifically infects Escherichia coli. Omnii modified the virus’s "legs," protein structures the phage uses to attach to bacterial cells, enabling these new variants to overcome the bacteria's defenses that had resisted existing antibiotics and phage treatments.
Experts outside the project have expressed cautious optimism. Michael Baym, a microbiologist at Harvard University, noted that if the treatment proves successful, it could represent a significant advance in personalized medicine targeting drug-resistant infections. However, he and others urge caution until the full clinical outcomes are available and the novelty and effectiveness of the designed viruses can be thoroughly evaluated. Brandon Berryhill, an evolutionary microbiologist at Uppsala University, highlighted the complexity in predicting phage therapy outcomes and noted that success in the laboratory does not guarantee clinical effectiveness.
The duration of phage therapy can vary from days to weeks, with ongoing monitoring to ensure infections do not relapse. If successful, Radical Numerics hopes that AI could significantly speed up the process of developing custom phages, potentially reducing design and production times to days, compared to the lengthy efforts currently required.
This trial represents a pioneering step towards integrating AI with phage therapy, a field that aims to address the critical global health challenge posed by antibiotic-resistant bacterial infections. Further research and peer-reviewed results will be essential to determine the long-term efficacy and safety of AI-designed viral therapies.