Machine learning identifies 800,000+ antimicrobial peptides

Nature has always been a good place to look for antibiotics, as bacteria have evolved numerous antibacterial defences — often in the form of short proteins (peptides) that can disrupt bacterial cell membranes and other critical structures. Now, an international research team has used machine learning to search for antibiotics in a vast dataset containing the recorded genomes of tens of thousands of bacteria and other primitive organisms — and identified over 800,000 potential antibiotic compounds in the process.

The study was co-led by researchers at the University of Pennsylvania, including co-senior author Assistant Professor César de la Fuente. In recent years, de la Fuente and colleagues have pioneered AI-powered searches for antimicrobials, identifying preclinical candidates in the genomes of contemporary humans, extinct Neanderthals and Denisovans, woolly mammoths and hundreds of other organisms.

“AI in antibiotic discovery is now a reality and has significantly accelerated our ability to discover new candidate drugs,” de la Fuente said. “What once took years can now be achieved in hours using computers.”

For this new study, the researchers used a machine learning platform to sift through multiple public databases containing microbial genomic data. The analysis covered 87,920 genomes from specific microbes as well as 63,410 mixes of microbial genomes, or metagenomes, from environmental samples spanning diverse habitats around the planet.

This extensive exploration succeeded in identifying 863,498 candidate antimicrobial peptides, more than 90% of which had never been described before. The identified compounds originated from microbes living in a wide variety of habitats — including human saliva, pig guts, soil and plants, corals, and many other terrestrial and marine organisms — validating the team’s broad approach to exploring the world’s biological data.

To validate their findings, the researchers synthesised 100 of the peptides and tested them against 11 disease-causing bacterial strains, including antibiotic-resistant E. coli and Staphylococcus aureus. They found that 79 disrupted the protective bacterial membranes and 63 completely eradicated the growth of at least one of the pathogens tested, and often multiple strains.

“In some cases, these molecules were effective against bacteria at very low doses,” de la Fuente said. Promising results were also observed in preclinical mouse models, where treatment with some of the peptides produced results similar to the effects of polymyxin B — a commercially available antibiotic which is used to treat meningitis, pneumonia, sepsis and urinary tract infections.

The team’s repository of putative antimicrobial sequences, which they call AMPSphere, is freely available to access at https://ampsphere.big-data-biology.org/. Their research has also been published in the journal Cell.

Image credit: iStock.com/inkoly