Recent News 19 : Reprogramming the War Against Superbugs: How AI-Powered Phage Therapy Is Reinventing Precision Antimicrobials

Reprogramming the War Against Superbugs: How AI-Powered Phage Therapy Is Reinventing Precision Antimicrobials

Antibiotic resistance has reached crisis levels globally. The overuse and misuse of antibiotics across human and veterinary medicine have driven bacteria to evolve rapidly, rendering many conventional drugs ineffective. According to the Global Research on Antimicrobial Resistance Project, antimicrobial resistance (AMR) has caused an estimated 1 million deaths annually since 1990, a number projected to climb as pathogenic bacteria continue to outpace pharmaceutical innovation.

In response to this looming threat, scientists and clinicians are revisiting one of microbiology’s most promising yet underutilized tools: bacteriophages, or phages. These viruses, which exclusively infect and destroy bacteria, offer a biological alternative to chemical antibiotics. Yet their full potential remains largely unrealized — not due to a lack of efficacy, but because of the immense complexity in selecting the right phage for a given bacterial infection. The process of identifying and validating phage-bacteria matches has historically depended on empirical methods that are time-consuming and resource-intensive.

A breakthrough in this domain is emerging from Paris, where Phagos, a tech-bio startup founded by microbiologist Adèle James and economist Alexandros Pantalis, is combining molecular microbiology, synthetic biology, and cloud-based artificial intelligence to scale precision phage therapy in real-world settings. Their mission: to create a new class of antimicrobials that are natural, resistance-proof, and adaptive, capable of fighting pathogens without the broad-spectrum collateral damage caused by traditional antibiotics.

Artistic view

Phages: Targeted Predators in a Sea of Generalist Antibiotics

Phages are the most abundant biological entities on Earth, with estimates placing their global population at over 10³¹ particles — that’s roughly a trillion phages for every grain of sand. Unlike antibiotics, which indiscriminately kill both pathogenic and beneficial bacteria, phages are highly specific, often infecting only a single bacterial species or strain. This precision is both their strength and their primary challenge.

“An antibiotic is like a bomb,” explains Pantalis. “It destroys everything in its path. A phage, on the other hand, is a scalpel.” But the specificity that makes phages so safe also makes them difficult to use. Identifying an effective phage against a clinical isolate has traditionally required laborious trial-and-error procedures, as there are potentially billions of unique combinations between phages and bacterial strains.

This inherent complexity creates a combinatorial bottleneck: every new infection, every new strain, and every new host context requires careful matching. The scale of the problem is not just large — it is exponential. If phage therapy is to be deployed widely in clinical or agricultural settings, a paradigm shift in how phages are discovered, validated, and deployed is necessary.

AI and the Genomic Revolution in Microbial Matchmaking

The solution proposed by Phagos is rooted in machine learning — more specifically, in generative artificial intelligence. Using both publicly available datasets and experimental results generated in their own laboratory, Phagos has developed a platform that predicts phage-host compatibility from genomic data. The system ingests data on bacterial surface receptors, known resistance mechanisms, CRISPR spacer histories, and environmental metadata to determine the likelihood of successful infection and lysis by specific phages.

This approach transforms phage selection from a hands-on, reactive task into a proactive, data-driven prediction engine. Rather than physically screening thousands of phage candidates against every clinical isolate, Phagos can now simulate these interactions computationally — a process that takes hours instead of weeks and scales with minimal marginal cost.

Their AI model has effectively become a computational microscope, capable of analyzing the hidden dynamics between phage and host at a molecular level. This leap in capability is due in large part to their partnership with Amazon Web Services (AWS).

AWS as Infrastructure for a Biological Intelligence Platform

In 2024, Phagos was selected for the AWS Generative AI Accelerator Program, which provided cloud credits, mentorship, and most critically, embedded AWS experts in machine learning and infrastructure. These full-time engineers worked alongside Phagos’s scientific team to build, optimize, and deploy large-scale bioinformatics models that are now the backbone of the company’s predictive capabilities.

This partnership allowed Phagos to transition from a wet-lab startup to a hybrid AI–microbiology platform, capable of not just processing biological data but learning from it continuously. Every new phage tested, every new bacterial genome sequenced, feeds back into the model, improving its predictive accuracy and expanding its knowledge base.

“We’re getting closer to being able to read into the DNA and determine whether a phage will work — before even touching a Petri dish,” said Pantalis. This represents a fundamental shift in microbiological methodology: from reactive empiricism to proactive computational biology.

From Oysters to Livestock: Early Applications in Animal Health

While the long-term goal of Phagos is human healthcare, the company has begun by targeting animal agriculture, which is responsible for the majority of global antibiotic usage. The team’s first proof of concept came from an oyster farm in France, where phage application led to a 40% reduction in mortality due to bacterial infections — all achieved with just $4,000 worth of lab equipment and a bench rented in a shared facility.

From this modest success, the platform has been expanded to target infections in shrimp, poultry, swine, and cattle, using phage cocktails dissolved in drinking water as a simple, scalable delivery method. This approach sidesteps many of the regulatory and logistical challenges faced by injectable or topical formulations and aligns with current industrial farming practices.

This isn’t just about animal welfare — it’s about breaking the chain of resistance. Antibiotic-resistant bacteria in farm animals do not remain confined to barns or aquaculture tanks; they migrate, share resistance genes, and ultimately infiltrate human environments. Reducing antibiotic use in agriculture is a critical pillar in the broader fight against AMR, and phage therapy represents one of the few interventions capable of achieving that at scale.

A New Identity: From Biotech to AI Model Builder

What began as a microbiology venture is now evolving into something much broader: a model builder. As Pantalis notes, “A quarter of the company is now focused entirely on data and machine learning.” This marks a significant departure from the typical profile of a biotech firm, underscoring how computational biology is no longer an auxiliary function but a core driver of discovery.

The infrastructure built with AWS is not just a support system — it is the engine of innovation. Every phage cocktail deployed feeds new information into the platform, refining its models. With each iteration, the system becomes more accurate, more autonomous, and more scalable. This feedback loop positions Phagos at the frontier of a new kind of medicine — one where biological agents evolve alongside pathogens, guided by a continuously learning algorithm.

Toward the Future: Adaptive Antimicrobials at Scale

Phagos’s ultimate ambition is to bring phage therapy into mainstream clinical practice by 2030, not as a niche or experimental treatment but as a core therapeutic modality alongside — or even replacing — antibiotics. Unlike fixed-spectrum drugs, phages are biological intelligences that co-evolve with their targets. When integrated with AI prediction systems, they become living therapeutics capable of adapting faster than bacteria can resist.

In contrast to companies chasing static, off-the-shelf phage cocktails, Phagos is betting on personalized therapy at scale: adaptive, precise, and continuously improving. This vision represents a paradigm shift in how we think about antimicrobials — not as chemical weapons, but as biological allies, guided by machine intelligence.

The war against superbugs will not be won by reloading old ammunition. It requires a new arsenal, a new strategy, and new alliances between disciplines. In Phagos’s case, it is the unlikely fusion of Parisian lab benches, generative AI, and cloud computing infrastructure that might just give humanity its next great weapon in the microbial arms race.

References :

1. Murray, C. J. L., et al. (2022). "Global burden of bacterial antimicrobial resistance in 2019: a systematic analysis." The Lancet, 399(10325), 629-655.

2. Clokie, M. R. J., et al. (2011). "Phages in nature." Bacteriophage, 1(1), 31–45.

3. Hatfull, G. F. (2020). "Actinobacteriophage research: expanding our understanding of bacterial viruses." Annual Review of Virology, 7, 37–58.

4. Phagos. (2025). Company website and AWS Accelerator Program documentation.

5. Global Research on Antimicrobial Resistance (GRAM), University of Oxford and the Institute for Health Metrics and Evaluation (IHME). (2022).