Researchers at the University of Pennsylvania have identified a new class of potential antibiotics hidden within prions, the misfolded proteins typically associated with fatal neurodegenerative conditions. By using the deep-learning platform APEX 1.1 to scan 19.3 million protein fragments, the team discovered 1,179 antimicrobial candidates—dubbed “prionins”—that can kill drug-resistant bacteria, according to findings published in Nature Microbiology.
How AI Unlocked Hidden Antibiotics
The discovery process relied on the ability of artificial intelligence to identify functional sequences that traditional laboratory screening often misses. César de la Fuente, PhD, director of the Machine Biology Group at the University of Pennsylvania, explains that the team utilized APEX 1.1 to analyze 2,897 prion and prion-like proteins. This process isolated 1,179 “prionins,” which are short peptide fragments capable of neutralizing pathogens, according to the study.
The team tested 75 of these peptides in the lab. Of those, 59 successfully inhibited at least one bacterial pathogen, and 42 showed high potency at low concentrations, a key metric for antibiotic effectiveness.
Testing Prionins Against Drug-Resistant Bacteria
To move beyond computer modeling, the researchers conducted experiments on both cells and animal models. According to co-first author Marcelo D. T. Torres, the team verified that many of these molecules function by disrupting bacterial membranes, a common strategy for antimicrobial peptides. In a controlled mouse model, researchers applied these peptides to skin infections caused by Acinetobacter baumannii. The treatment reduced bacterial levels comparable to the antibiotic polymyxin B, with no observed weight loss or toxicity in the subjects, according to the study data.
Why This Changes Antibiotic Discovery
Historically, drug discovery has been restricted by human bias regarding which proteins are worth investigating. While prions are primarily studied for their role in neurodegeneration, this research suggests they contain “encrypted peptides” that serve as a natural defense mechanism. This approach contrasts with traditional methods that often focus on well-documented antimicrobial sources like venoms or common bacterial secretions. By mining the “hidden layers” of proteins, the Penn team is expanding the search space for new treatments at a time when antibiotic resistance is increasingly limiting clinical options, according to the researchers.
Pro Tip: The Power of Encrypted Peptides
Researchers are increasingly looking at “encrypted peptides”—short, functional sequences hidden within larger proteins. If you are tracking biotech trends, watch for studies that use machine learning to “unlock” these sequences from previously ignored biological sources, such as extinct organisms or human waste products.
Frequently Asked Questions
Are these prion-based antibiotics dangerous?
No. The study indicates that the “prionins” identified are fragments of proteins, not the misfolded, infectious prions themselves. Researchers tested 16 active peptides and found no measurable harm to human red blood cells or other cells, according to the study.
Will these treatments replace current antibiotics?
The research is currently in the experimental stage. While the results in mice are promising, these candidates must undergo further clinical trials to determine their safety and efficacy in humans, according to the University of Pennsylvania.
What are “prionins”?
Prionins are a newly identified class of short antimicrobial peptides found within prion and prion-like proteins. They were named by the University of Pennsylvania research team after they were identified using the APEX 1.1 deep-learning platform.
Are you interested in the intersection of AI and modern medicine? Subscribe to our newsletter for the latest updates on how machine learning is reshaping drug discovery. Have a question about this research? Leave a comment below.
