Discovery of Antimicrobial Lysins from the “Dark Matter” of Uncharacterized Phages Using Artificial Intelligence

• Published in: Advanced science Vol. 11; no. 32; pp. e2404049 - n/a
• Main Authors: Zhang, Yue, Li, Runze, Zou, Geng, Guo, Yating, Wu, Renwei, Zhou, Yang, Chen, Huanchun, Zhou, Rui, Lavigne, Rob, Bergen, Phillip J., Li, Jian, Li, Jinquan
• Format: Journal Article
• Language: English
• Published: Germany John Wiley & Sons, Inc 01.08.2024; John Wiley and Sons Inc; Wiley
• Subjects: antibacterial protein; antibiotic resistance; Antibiotics; Artificial intelligence; Bacteria; Bacterial infections; Dark matter; Drug resistance; Genomes; high‐throughput screening; infectious diseases; Kinases; Machine learning; Mining; Pathogens; phage lysin; prophage; Proteins; Software; stacking model; phage lysin; prophage; stacking model; antibiotic resistance; high‐throughput screening; infectious diseases; antibacterial protein
• ISSN: 2198-3844; 2198-3844
• DOI: 10.1002/advs.202404049

The rapid rise of antibiotic resistance and slow discovery of new antibiotics have threatened global health. While novel phage lysins have emerged as potential antibacterial agents, experimental screening methods for novel lysins pose significant challenges due to the enormous workload. Here, the first unified software package, namely DeepLysin, is developed to employ artificial intelligence for mining the vast genome reservoirs (“dark matter”) for novel antibacterial phage lysins. Putative lysins are computationally screened from uncharacterized Staphylococcus aureus phages and 17 novel lysins are randomly selected for experimental validation. Seven candidates exhibit excellent in vitro antibacterial activity, with LLysSA9 exceeding that of the best‐in‐class alternative. The efficacy of LLysSA9 is further demonstrated in mouse bloodstream and wound infection models. Therefore, this study demonstrates the potential of integrating computational and experimental approaches to expedite the discovery of new antibacterial proteins for combating increasing antimicrobial resistance. DeepLysin is the first unified software that employs artificial intelligence to mine novel antibacterial proteins from the “dark matter.” The best‐in‐class antibacterial protein LLysSA9 is identified, which can effectively kill drug‐resistant bacteria in vitro and in vivo, being a potent candidate with clinical application prospects. This software expedites the discovery of new antibacterial proteins for combating increasing antimicrobial resistance.