In the 20th century, antibiotics transformed medicine. Infections that once killed millions could be cured with a pill or injection. Surgeries became safer, cancer treatments more effective, and advanced medical interventions, such as organ transplants, became possible, all because doctors could rely on these drugs to control infections. Unfortunately, today, that foundation is crumbling. Bacteria are evolving faster than medicine can keep up. Common antibiotics are failing, and infections that were once easily treatable are becoming deadly again. In 2019 alone, antimicrobial resistance was linked to nearly five million deaths worldwide, making it deadlier than HIV or malaria. The economic cost is equally staggering: the World Bank warns of trillions lost in global productivity and millions pushed into poverty if nothing changes. This crisis, caused by antimicrobial resistance, has been described as a “silent pandemic.” Unlike a sudden outbreak, it spreads quietly, making routine medical care slightly more dangerous each year. Yet amid this grim outlook, new research is opening a window of hope. At the forefront of new innovations in this area are Dr. Kai Hilpert of City St George’s, University of London, and his colleagues, who are pioneering an approach that combines biology, chemistry, and artificial intelligence to reinvent how we discover infection-fighting medicines. Their work has been recognised with a prestigious award from the UK’s Biotechnology and Biological Sciences Research Council, BBSRC. More
So, who are the people behind this project? Kai Hilpert has been pioneering the use of Spot peptide synthesis with artificial intelligence since 2007, creating and testing over ten thousand antimicrobial peptides. Inanc Birol, in Vancouver, is a leader in computational biology who co-developed the deep-learning tool AMPlify, bringing the power of high-performance computing to predict new antimicrobial peptides. Caren Helbing, at the University of Victoria in Canada, is an expert in genomics and proteomics, and her work with Birol on AMPlify adds critical biological depth to link predictions with real-world health applications. Jayachandran Kizhakkedathu, also from Vancouver, is an expert in polymer chemistry and biomaterials, contributing unique expertise in designing infection-resistant materials and developing ways to make peptides effective in biomedical settings. And in Germany, Ralf Mikut is a machine learning specialist who created the SciXMiner analytics platform and ensures peptide data and AI models are integrated, interpreted, and applied effectively.
Long before humans learned to isolate penicillin from a fungus, nature had already equipped living creatures with a defence system against microbes: antimicrobial peptides, or AMPs for short. These tiny chains of amino acids are found in everything from frogs and insects to plants and humans. They serve as miniature shields, punching holes in bacterial membranes, disrupting their inner workings, or alerting the immune system to fight harder.
AMPs are attractive candidates for new medicines to treat infections because bacteria struggle to resist them. Unlike traditional antibiotics that usually target one key protein, AMPs attack on multiple fronts at once. That makes it far harder for microbes to develop a simple genetic trick to survive.
But AMPs have weaknesses too. Many are unstable in the bloodstream, breaking down before they reach their target. Some can also harm human cells, and making them in large quantities is expensive. For decades, these hurdles kept AMPs in the realm of theory rather than routine therapy.
That’s where Dr. Hilpert and colleagues stepped in. Their project, called AMPQuest, was built around a simple but bold idea: use artificial intelligence to design better AMPs, and then test them under conditions that better mimic the human body.
Hilpert’s team trained powerful computer models to predict which amino acid combinations might work. The AI system used, called AMPlify, can scan vast sequence databases and highlight promising candidates.
Once predicted, the team used a clever laboratory method called SPOT synthesis to produce thousands of peptide candidates quickly. These candidates were then tested not just in simple broth (a simple medium used to culture bacteria), but also in broth with 20% human serum added, an environment much closer to the bloodstream, where many promising drugs usually fail.
This careful step matters. Many earlier AMP experiments produced peptides that looked strong in the lab but lost all potency in real biological conditions. By raising the bar from the start, Hilpert and his colleagues avoided wasting time on molecules that would never survive inside the body.
The first round of experiments was sobering: while more than two-thirds of the AI-designed peptides killed bacteria in broth, fewer than six percent remained active in serum. It was a reminder of just how challenging the real-world physiological environment is.
Cunningly, the team fed the first-round results back into their AI model, essentially teaching it what worked and what didn’t. Strikingly, In the next round, nearly 40 percent of peptides retained their strength in serum, a sevenfold improvement.
Among these, about twenty peptides stood out. They killed multidrug-resistant E. coli, one of the most dangerous hospital pathogens, while showing no signs of toxicity in human cells. Remarkably, some even became more effective in serum than in broth, making them up to 5,000-fold more potent than typically observed for AMPs tested under these conditions.
For researchers who have long struggled with the fragility of AMPs, this was a very exciting result. It suggested that with the right blend of computation and careful testing, the dream of peptide-based antibiotics might finally be within reach.
The implications of these results are significant. If further studies confirm safety and effectiveness, these peptides could form the basis of entirely new medicines, ones that bacteria would find far harder to resist. They could safeguard surgeries, protect cancer patients, and save newborns from infections that today carry terrifying risks.
Dr. Hilpert and colleagues are cautious, acknowledging that many steps remain. The exact mechanisms of these peptides need to be investigated and understood to ensure their efficacy and safety in human patients. Their performance must be tested against a wider variety of pathogens, not just E. coli. More extensive safety studies are essential. And eventually, animal and human trials will determine whether these laboratory successes translate into real cures.
However, the antimicrobial drug discovery pathway looks more promising than it has in years. The combination of artificial intelligence, high-speed synthesis, and rigorous testing represents a new model for antibiotic discovery, one that could be scaled up to tackle not just bacterial infections, but fungal and viral threats too.
Traditional antibiotic discovery has slowed to a crawl, with pharmaceutical companies reluctant to invest in costly projects that may not yield profits. By contrast, projects such as Hilpert’s show how academic and international collaboration, powered by modern computational tools, can reopen paths that once seemed closed.
It also reflects a larger truth: the fight against antimicrobial resistance will not be won by one silver bullet. It requires many strategies, such as responsible antibiotic use, improved hygiene, vaccination, and alternatives such as peptides. But each successful step matters, and each new lead compound buys time for humanity to stay ahead in this evolutionary arms race.
The peptides Dr. Hilpert and his international colleagues are uncovering may one day protect patients around the world. Their work offers proof that innovation is still possible, even in the face of one of the greatest medical challenges of our time.
