Image by freepik

Almost a century ago, the discovery of antibiotics like penicillin revolutionized medicine by harnessing the natural bacteria-killing abilities of microbes. However, with the overuse of antibiotics, more and more bacteria have developed resistance, becoming "superbugs" and putting humanity back under the threat of microbes. Antibiotic-resistant infections are becoming increasingly difficult to treat with traditional methods. According to data from the World Health Organization, currently, 1.27 million people die each year from antibiotic-resistant infections, making it one of the most severe public health threats.

Due to the fact that many past antibiotics were derived from soil microorganisms, developing new antibiotics using traditional methods is not easy. In recent decades, very few new types of antibiotics have been created, and their structures are largely similar to those of existing ones. Humanity urgently needs new methods for discovering antibiotics.

Fortunately, with the rapid development of artificial intelligence (AI), humanity can now accelerate the identification of new antibiotics. In a recent paper published in the top academic journal Cell, scientists from institutions such as Queensland University of Technology (QUT) in Australia, Fudan University's Institute of Brain-Inspired Intelligence in China, and the University of Pennsylvania in the United States collaborated to use machine learning. They enabled AI to search through large microbial genome datasets from around the world and identified 863,498 promising antimicrobial peptides (AMPs).

Antimicrobial peptides are short peptide molecules composed of 10 to 100 amino acids that can kill or inhibit the growth of infectious microorganisms. Many bacteria in nature, such as Staphylococcus, Vibrio cholerae, and Listeria, produce antimicrobial peptides to eliminate competitors and gain an advantageous position in their ecological niches.

Natural antimicrobial peptides can be produced through protein decomposition, non-ribosomal synthesis, or encoded within the genome. In this study, scientists focused on the latter method. The research team screened metagenomes, which are microbial genome mixtures extracted from over 60,000 environmental samples worldwide, including samples from oceans, soil, plants, human saliva, and animal intestines.

Researchers established a massive dataset encompassing the genomes of tens of thousands of bacteria and other primitive organisms. Through extensive exploration, they successfully identified nearly a million candidate antimicrobial peptide sequences, over 90% of which had not been previously described.

Diagram illustrating the screening of antimicrobial peptide sequences from global microbial genome samples.

Credit: Cell(2024),DOI:10.1016/j.cell.2024.05.013

To validate the machine's predictions, the research group synthesized 100 antimicrobial peptides in the lab and tested them both in vitro and in vivo against drug-resistant pathogens that cause clinical infections and human gut symbiotic bacteria. The tests revealed that 79 out of these 100 peptides exhibited antibacterial activity capable of disrupting bacterial membranes. Of these, 63 peptides specifically inhibited the growth of a single pathogen entirely, including antibiotic-resistant bacteria like Staphylococcus aureus and Escherichia coli.

Tests on mouse models infected with bacteria showed that some antimicrobial peptides had anti-infection effects comparable to commercial polymyxin B—a polypeptide antibiotic approved for treating meningitis, pneumonia, sepsis, and urinary tract infections.

Anti-infective activity of AMPs in preclinical animal model

Credit: Cell(2024),DOI:10.1016/j.cell.2024.05.013

These initial tests against pathogenic bacteria indicate that the new antimicrobial peptide database could significantly enhance human strategies for combating bacterial infection threats.

Professor César de la Fuente, one of the main authors of the study, stated: "AI in antibiotic discovery is now a reality and has significantly accelerated our ability to discover new candidate drugs. What once took years can now be achieved in hours using computers"

The research team has made the AI-identified antimicrobial peptide sequences publicly available for free access: https://ampsphere.big-data-biology.org/

References

1. Santos-Júnior CD, Torres MDT, Duan Y, et al. Discovery of antimicrobial peptides in the global microbiome with machine learning. Cell. Published online May 30, 2024. doi:10.1016/j.cell.2024.05.013.

2. University of Pennsylvania School of Medicine. "Largest-ever antibiotic discovery effort uses AI to uncover potential cures in microbial dark matter." ScienceDaily. www.sciencedaily.com/releases/2024/06/240605162425.htm (accessed June 7, 2024).