AI Unveils Antibiotic Potential in Prion Proteins

Researchers at the University of Pennsylvania’s Perelman School of Medicine have identified a new class of antibiotic candidates called “prionins” hidden within proteins usually linked to fatal brain diseases. Using the AI platform APEX 1.1, the team discovered these antimicrobial peptides can kill drug-resistant bacteria, according to a study published in Nature Microbiology.

How did AI discover “prionins” in brain proteins?

The research team used a deep-learning platform called APEX 1.1 to scan 2,897 prion and prion-like proteins. This process analyzed 19.3 million short peptide fragments to predict antibiotic activity.

The AI identified 1,179 candidate antimicrobial peptides, which the researchers named “prionins.” César de la Fuente, PhD, FRSB, director of the Machine Biology Group at the Perelman School of Medicine, stated that AI allowed the team to ask if these proteins encode useful molecular fragments rather than seeing them only through the lens of disease.

Did You Know? The AI search analyzed 19.3 million short peptide fragments across nearly 2,900 different prion and prion-like proteins to find these candidates.

Do these new antibiotic candidates work in living subjects?

Researchers tested 75 of the most promising peptides against 11 bacterial pathogens, including drug-resistant strains. Fifty-nine of these inhibited at least one pathogen, and 42 showed strong activity at low concentrations.

In mouse models, two peptides—one derived from a roundworm and one from a fungus—reduced bacteria levels in skin infections caused by Acinetobacter baumannii. These results were comparable to the drug polymyxin B, and the mice showed no treatment-related weight loss.

Marcelo D. T. Torres, co-first author of the study, noted that the success of these molecules in the lab and animal models transforms the project from a prediction exercise into a discovery platform.

Expert Insight: Samantha Carter suggests that the ability to repurpose proteins associated with neurodegeneration for immune defense highlights a shift in how scientists view “disease” proteins. By isolating “encrypted peptides,” researchers may be able to unlock therapeutic tools from biological sources previously dismissed as purely harmful.

What is the biological significance of this discovery?

Many of the active prionins work by disrupting bacterial membranes, a common strategy for antimicrobial peptides. Toxicity tests showed that 16 active peptides caused no measurable harm to human cells or red blood cells at the highest concentrations tested.

Accelerating antibiotic discovery using AI | César de la Fuente

This work is part of a larger effort by the de la Fuente Lab to find “encrypted peptides” in various sources, including venoms, microbiomes, and extinct organisms. The study suggests a possible link between protein aggregation and host defense.

What happens next for prionin research?

The findings may lead to further exploration of whether prion-like proteins naturally act as antibiotics in the body. This could result in a new pipeline for developing treatments for infections that are currently difficult to treat.

Future research may also focus on identifying more encrypted peptides within other protein classes. While this does not change the known harmful role of misfolded prions in brain disease, it may expand where scientists look for new medicines.

Frequently Asked Questions

What are prionins?
Prionins are a new class of short antimicrobial peptides discovered by AI within prion and prion-like proteins.

Which bacteria did the prionins fight in animal tests?
The researchers used the peptides to treat skin infections caused by Acinetobacter baumannii, a difficult-to-treat pathogen, in mice.

Does this mean prions are no longer dangerous?
No. The researchers stated that this work does not change what is known about the harmful role of misfolded prions in neurodegenerative diseases.

Could AI-driven searches in other “disease” proteins reveal more hidden medical treatments?