AI Discovers New Antibiotics to Combat Drug-Resistant Bacteria

Researchers have harnessed the power of artificial intelligence (AI) to uncover a novel category of antibiotics capable of treating infections caused by drug-resistant bacteria. This breakthrough has the potential to address the significant issue of antibiotic resistance, which claimed over 1.2 million lives in 2019 and is projected to worsen in the coming years. The findings of this study were recently published in the prestigious scientific journal Nature.

The experimental tests conducted on mice have demonstrated the promising efficacy of a new antibiotic compound against two notorious drug-resistant bacterial strains: methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin-resistant enterococci. These bacteria are known to be highly resistant to commonly used drugs for MRSA infections.

Initially, researchers assessed over 39,000 compounds on Staphylococcus aureus and three types of human cells, including liver, skeletal muscle, and lung cells. These findings served as training data for the AI model, enabling it to learn the chemical composition and bonding patterns of each compound. Consequently, the AI model was able to predict the antimicrobial activity of the compounds, as well as their potential toxicity to human cells.

Leveraging this trained AI model, the researchers then conducted computer simulations on a staggering 12 million compounds. Through this process, they identified 3,646 compounds with desirable pharmaceutical properties. Further analysis allowed them to pinpoint specific chemical substructures that explained the properties of each compound.

By comparing these substructures across various compounds, the team identified entirely new classes of potential antibiotics. Ultimately, they discovered two non-toxic compounds capable of effectively eliminating MRSA and vancomycin-resistant enterococci.

To validate the effectiveness of these compounds, the researchers conducted experiments on mice, demonstrating their ability to treat skin and thigh infections caused by MRSA.

James Collins, an author of the study from the Broad Institute, emphasized the scarcity of new antibiotic classes capable of combating both MRSA and vancomycin-resistant enterococci. He noted that resistance to the few existing antibiotics in this regard, such as oxazolidinone and lipopeptide antibiotics, has been growing. Collins stated that their research had identified a new antibiotic—one of the few in the past 60 years—that could serve as a complement to existing antibiotics.

Encouraged by these results, researchers are now utilizing the AI-guided approach to design entirely novel antibiotics and explore other drug classes. This methodology holds promise for the discovery of compounds that can selectively target and eliminate aging or damaged cells implicated in various diseases, including osteoarthritis and cancer.