published late last week in the journal Nature Microbiology.

Good guy prions?

Prions are some of the strangest things around. They’re the misfolded form of a protein naturally found in the body. When a prion comes across its “normal” counterpart, it can somehow induce the latter to turn into a prion itself, almost like a zombie infection.

Classic prion disorders like mad cow disease and Creutzfeldt-Jakob disease are caused by the steady accumulation of one particular type of protein, aptly named the prion protein; these disorders are universally fatal. Some scientists have also argued that other neurodegenerative conditions, such as Alzheimer’s disease, are caused by other kinds of misfolded proteins that act in a similar way to prions.

According to the study researchers, there’s growing evidence that prions and prion-like proteins are more than just harbingers of death. Studies have found that the normal prion protein and the prion-like amyloid beta (one of the drivers of Alzheimer’s) can have antimicrobial activity, for instance. So the team decided to conduct a sweeping analysis looking for antimicrobial peptide fragments within these proteins.

The researchers had previously built an AI model intended to predict the antimicrobial activity of any given peptide fragment, named APEX 1.1. Then they let APEX scan through 19.3 million short peptide fragments found in 2,897 prion and prion-like proteins. They initially uncovered 1,179 candidates, which the team narrowed down to 75 that showed the most potential. Of these, 59 were able to inhibit the growth of at least one bacterial germ in the lab, including 42 that did so at low levels (important for dosing considerations).

Finally, the researchers tested two of the strongest candidates on the skin of mice infected with Acinetobacter baumannii, a common source of drug-resistant infections in people. The candidates appeared to be roughly as effective as polymyxin B, an existing antibiotic often used as a last resort drug for certain drug-resistant infections.

The researchers have coined these antibacterial fragments collected from prions as “prionins.”

The future of prionins

More research is obviously needed to verify whether the team’s prionins can actually work as hoped—and safely—in people. The researchers also note their findings don’t settle the open question as to whether prions or prion-like proteins naturally tackle bacterial infections in our body.

At the same time, they do argue their work provides a strong proof of concept that prionins identified through AI can be viable antibiotic candidates for further testing.

“For a long time, drug discovery has been limited not only by what we can test, but by where we choose to look,” said senior study author César de la Fuente, director of the Machine Biology Group at the University of Pennsylvania Perelman School of Medicine, in a statement from the university. “AI is changing that. It gives us a way to search the hidden layers of biology and ask whether molecules associated with one story—in this case, disease—may also carry another story with therapeutic potential.”

With any luck, the proteins known for causing the scariest diseases around could someday turn into our antibacterial allies.

The next frontier of antibiotics might come from an unexpected place. Recent research identifies potential antibiotic candidates from inside prions—proteins capable of causing some of the deadliest brain infections ever known, such as mad cow disease.

Scientists at the University of Pennsylvania used artificial intelligence to rapidly search hundreds of prions and prion-like proteins for peptides with antibacterial activity. They found several dozen promising candidates, two of which have already shown results treating bacterial infections in mice.

The team’s findings establish “prion-related proteins as a productive source space for antibiotic discovery,” the scientists wrote in their paper, published late last week in the journal Nature Microbiology.

Good guy prions?

Prions are some of the strangest things around. They’re the misfolded form of a protein naturally found in the body. When a prion comes across its “normal” counterpart, it can somehow induce the latter to turn into a prion itself, almost like a zombie infection.

Classic prion disorders like mad cow disease and Creutzfeldt-Jakob disease are caused by the steady accumulation of one particular type of protein, aptly named the prion protein; these disorders are universally fatal. Some scientists have also argued that other neurodegenerative conditions, such as Alzheimer’s disease, are caused by other kinds of misfolded proteins that act in a similar way to prions.

According to the study researchers, there’s growing evidence that prions and prion-like proteins are more than just harbingers of death. Studies have found that the normal prion protein and the prion-like amyloid beta (one of the drivers of Alzheimer’s) can have antimicrobial activity, for instance. So the team decided to conduct a sweeping analysis looking for antimicrobial peptide fragments within these proteins.

The researchers had previously built an AI model intended to predict the antimicrobial activity of any given peptide fragment, named APEX 1.1. Then they let APEX scan through 19.3 million short peptide fragments found in 2,897 prion and prion-like proteins. They initially uncovered 1,179 candidates, which the team narrowed down to 75 that showed the most potential. Of these, 59 were able to inhibit the growth of at least one bacterial germ in the lab, including 42 that did so at low levels (important for dosing considerations).

Finally, the researchers tested two of the strongest candidates on the skin of mice infected with Acinetobacter baumannii, a common source of drug-resistant infections in people. The candidates appeared to be roughly as effective as polymyxin B, an existing antibiotic often used as a last resort drug for certain drug-resistant infections.

The researchers have coined these antibacterial fragments collected from prions as “prionins.”

The future of prionins

More research is obviously needed to verify whether the team’s prionins can actually work as hoped—and safely—in people. The researchers also note their findings don’t settle the open question as to whether prions or prion-like proteins naturally tackle bacterial infections in our body.

At the same time, they do argue their work provides a strong proof of concept that prionins identified through AI can be viable antibiotic candidates for further testing.

“For a long time, drug discovery has been limited not only by what we can test, but by where we choose to look,” said senior study author César de la Fuente, director of the Machine Biology Group at the University of Pennsylvania Perelman School of Medicine, in a statement from the university. “AI is changing that. It gives us a way to search the hidden layers of biology and ask whether molecules associated with one story—in this case, disease—may also carry another story with therapeutic potential.”

With any luck, the proteins known for causing the scariest diseases around could someday turn into our antibacterial allies.