New antibiotic candidates against drug-resistant bacteria may exist in prions, misfolded proteins in the brain best known for rare and fatal degenerative brain diseases. A new study from the Perelman School of Medicine at the University of Pennsylvania suggests that prions and prion-like proteins may harbor short peptides called prionins that can kill bacteria, and that the protein best known for its role in neurodegeneration may contain molecular signatures associated with immune defense.
From deadly brain diseases to the discovery of antibiotics
The survey results announced today are natural microbiologypresents a surprising new place to look for antibiotic candidates at a time when drug-resistant bacterial infections are narrowing treatment options. The study also raises a broader biological question: whether the proteins most frequently associated with neurodegeneration contain hidden molecular signatures associated with innate immunity.
Previous research has suggested this association. Researchers had previously reported that fragments of some proteins, such as amyloid beta and cellular prion protein, which are involved in neurodegenerative diseases such as Alzheimer’s disease, may fight microbes. However, no one had conducted a systematic large-scale search of prions and prion-like proteins for hidden antimicrobial peptides. Penn’s team used AI to do it.
AI search reveals hidden classes of antimicrobial peptides
The University of Pennsylvania team used a deep learning platform called APEX 1.1 to scan 19.3 million short peptide fragments from 2,897 prions and prion-like proteins. APEX predicts the antibiotic activity of specific amino acid sequences and identifies 1,179 candidate antimicrobial peptides. The researchers named this new class “prionins.”
This research changes where antibiotics are thought to be hiding. Prions have long been viewed almost entirely through the lens of disease, but AI allows us to ask another question: whether these proteins also encode useful molecular fragments. The answer seems to be yes. ”
Dr. Cesar de la Fuente, FRSB, Presidential Associate Professor and Director of the Mechanical Biology Group at the Perelman School of Medicine at the University of Pennsylvania, and senior author of the study
Validate promising candidates in lab and mouse tests
The research team selected 75 of the most promising peptides for experimental testing based on the platform’s assessment of how well they performed against 11 different bacterial pathogens, including drug-resistant strains. Of them, 59 inhibit at least one bacterial pathogen, and 42 exhibit strong activity at low concentrations, a particularly important designation.
Additional experiments suggest that many of the active prionins work by disrupting bacterial membranes, a common strategy used by antimicrobial peptides. Signs of toxicity were limited, and the 16 active peptides showed no measurable harm to red blood cells or human cells, even at the highest concentrations tested.
To test these findings, the researchers tested two of the most promising peptides, one from a fungus and one from a roundworm, in mice. They found that this approach reduced bacterial levels in standard skin infection models caused by: Acinetobacter baumanniiis a pathogen that is difficult to treat. Its efficacy was comparable to polymyxin B, and the researchers found no treatment-related weight loss.
“Here, the story becomes more than just a computer screen,” said Marcelo DT Torres, co-lead author of the study. “The AI search yielded a short list of candidates, but importantly, many of those molecules worked in the lab and two worked in animal infection models. This is what makes this a discovery platform and not just a predictive exercise.”
A new frontier in antibiotic discovery
The discovery builds on the de la Fuente Institute’s extensive efforts to unearth “cryptopeptides” from the biological world – short sequences hidden within larger proteins that may have biological functions when isolated. The group’s previous research has investigated human proteins, extinct organisms, archaea, the microbiome, and venoms. Prion research extends that idea to one of biology’s most unexpected classes of proteins.
This study also raises interesting possibilities at the intersection of neurodegeneration and innate immunity. This does not indicate that prionin is naturally released during infection or that prions and prion-like proteins normally act as antibiotics in the body. It also does not change what is known about the deleterious role of misfolded prions in neurodegenerative diseases. Rather, this study suggests that these proteins may be a rich source of previously overlooked antibiotic candidates and a new place to ask questions about the link between protein aggregation and host defense.
“For a long time, drug discovery has been limited not only by what you can test, but also where you can test it,” de la Fuente said. “AI is changing that. It provides a way to search the hidden layers of biology and ask whether molecules associated with one story (in this case, a disease) are also responsible for another story with therapeutic potential.”
sauce:
University of Pennsylvania School of Medicine
Reference magazines:
Torres, MDT; Others. (2026). Deep learning reveals antimicrobial peptides within prions. natural microbiology. DOI: 10.1038/s41564-026-02408-1. https://www.nature.com/articles/s41564-026-02408-1
