Discovery of Archaeal Antimicrobial peptides using deep learning

( 12.26.25  )                                    Written by Bowen Zhang            (special edition)        

Antibiotic resistance is a growing health concern, as many bacterial pathogens are becoming increasingly resistant to existing treatments. Today, most antibiotics have been discovered through bacteria or fungi, while archaea’s potential as a discovery pipeline for antimicrobials has been largely unexplored. In this study, researchers investigated whether or not archaeal proteins could serve as a new source of antibiotics by using deep learning to analyze the archaeal proteome.

The researchers developed a computational model to scan the protein sequences from hundreds of archaea species to identify short peptide fragments predicted to have antimicrobial activity. Using this approach, the researchers discovered thousands of potential peptide candidates, labelling them with the term “archaeasins”. A subset of these peptides was chemically synthesized and tested against clinically relevant bacteria pathogens to continue to evaluate their effectiveness.

Experimental results revealed that many archaea were capable of killing bacteria under laboratory conditions, including drug-resistant strains. Further testing revealed the peptides primarily work by disrupting the cell membrane, leading to cell death. Most notably, however, was the fact that several archaeasins demonstrated strong antibacterial capability while showing no harm towards mammalian cells.

This discovery reveals a new gateway of antimicrobial compound sources within the archaeal domain, along with highlighting the power of combining artificial intelligence with experimental microbiology. By expanding antibiotic discovery beyond traditional bacteria or fungi, this research offers a promising strategy for developing new treatments to combat antibiotic resistant infections in the future.

---------------------------------------------------------------------------------------------------------------------

You can read the full paper here:

https://www.nature.com/articles/s41564-025-02061-0