Extracted from the article: https://www.immedicohospitalario.es/noticia/48858/el-uso-de-algoritmos-predictivos-en-la-terapia-con-fagos-allana-el-cam.html
The design of an artificial intelligence program based on bacterial genome analysis enables the revival of bacteriophage therapy, which was first developed in the early 20th century. The goal is to use this type of virus—which only infects bacteria and specifically eliminates those that are pathogenic—as an alternative treatment for infections that are increasingly difficult to treat due to antimicrobial resistance (AMR).
December 3, 2024
The current rise in antibiotic resistance among human bacteria has led to renewed interest in phages as a therapeutic alternative known as phage therapy. Bacteriophages are viruses that infect and parasitize bacteria. Because they are active agents, these phages must undergo strict quality control to ensure the absence of undesirable effects.
Phage therapy was invented by Pasteur Institute scientist Félix d’Hérelle in the 1920s but was progressively abandoned after the late 1930s with the advent of antibiotics. Today, only a few Eastern European countries, such as Georgia, still use phage therapy. In most other countries, «broad-host-range» phages are used only occasionally and under compassionate use to treat chronic infections resistant to multiple drugs. One of the challenges in this field is determining which phage will be effective for a specific infection, as each phage can only infect certain bacterial strains.
In this context, scientists from the Pasteur Institute, Inserm, the Public Hospitals Network of Paris (AP-HP), and the University of Paris Cité (France) have announced the development of a new simple and effective tool that recommends the best possible phage cocktail. The idea is to use these viruses, which only infect bacteria, to specifically eliminate those that are pathogenic to humans, according to a study published in Nature Microbiology.
“We studied 350,000 interactions and managed to identify the characteristics of the bacterial genome that can predict phage efficacy,” summarizes Aude Bernheim, the study’s senior author and head of the Microbial Molecular Diversity Laboratory at the Pasteur Institute.
Florian Tesson, co-author and PhD candidate at the same laboratory and the IAME lab of University Paris Cité-Inserm, explained: “Contrary to what we initially thought, the phages’ ability to infect bacteria—which determines their effectiveness—is dictated by the bacterial surface receptors, not by the bacteria’s antiviral defense mechanisms.”
Bacterial genome and AI
This precise and comprehensive analysis of the interaction mechanisms between bacteria and phages enabled the team’s bioinformaticians to design an optimized and effective artificial intelligence program. The program is based on bacterial genome analysis, particularly in regions involved in coding bacterial membrane receptors—the entry point for phages.
After more than two years of development and training, the AI model was able to correctly predict phage efficacy in treating E. coli in 85% of cases in the dataset, simply by analyzing bacterial DNA. “This result exceeded our expectations,” says Aude Bernheim.
To further their research, the scientists tested the model on a new collection of E. coli strains responsible for pneumonia and selected a personalized “cocktail” of three phages for each. In 90% of cases, the phages chosen specifically by the AI were successful in destroying the bacteria.
In summary, this work provides quantitative insights into phage-host specificity and supports the use of predictive algorithms in phage therapy.
According to the authors, the method—which can be easily implemented in hospital laboratories—paves the way for personalized phage treatments in the coming years, especially in cases where Escherichia coli infections highly resistant to antibiotics are diagnosed.
“We still need to test phage effects in different environments, but a proof of concept has already been established. We hope to extend this to other pathogenic bacteria, since our AI model was designed to be easily adapted to other scenarios, with the goal of offering personalized phage therapy treatments in the future,” concludes Aude Bernheim.
