Researchers at the Perelman School of Medicine at the University of Pennsylvania have identified a novel class of antimicrobial peptides embedded within prion and prion-like proteins, challenging the conventional view of these molecules solely as agents of neurodegeneration. Published in Nature Microbiology, the study leverages artificial intelligence to uncover potential antibiotic candidates hidden within the biological sequences of proteins linked to fatal brain disorders. Led by César de la Fuente, director of the Machine Biology Group, the team utilized a deep-learning platform named APEX 1.1 to systematically analyze encrypted sequences within larger proteins. The AI screened 19.3 million short peptide fragments derived from 2,897 prion and prion-like proteins, predicting antibiotic activity based on amino acid sequences. This computational approach identified 1,179 candidate antimicrobial peptides, designated as prionins. To validate these predictions, researchers experimentally tested 75 of the most promising candidates against 11 distinct bacterial pathogens, including multidrug-resistant strains. Results indicated that 59 peptides inhibited at least one pathogen, with 42 demonstrating potent activity at low concentrations. Mechanistic assays suggested that these prionins function primarily by disrupting bacterial cell membranes. Notably, the candidates exhibited low toxicity; 16 active peptides showed no measurable harm to human or red blood cells at maximum tested concentrations. In vivo validation was conducted using murine models of skin infections caused by Acinetobacter baumannii, a recalcitrant pathogen. Two leading candidates, isolated from fungal and nematode sources, significantly reduced bacterial load with efficacy comparable to the established antibiotic polymyxin B, without causing treatment-related weight loss. This work expands the de la Fuente Lab's broader initiative to mine diverse biological sources for hidden therapeutic sequences. While the study does not establish that prionins function as natural antibiotics within the body or alter the understanding of prions' role in neurodegenerative disease, it highlights a significant opportunity in antibiotic discovery. The findings suggest that proteins associated with protein aggregation may harbor molecular features relevant to innate immunity, prompting new biological inquiries. By shifting the search space for drug candidates to previously overlooked protein classes, this research underscores the transformative potential of AI in identifying therapeutic molecules embedded within the hidden layers of biology.