Researchers at the Massachusetts Institute of Technology have used generative artificial-intelligence models to design two structurally novel antibiotic candidates that combat two of the world’s most stubborn drug-resistant pathogens: Neisseria gonorrhoeae and methicillin-resistant Staphylococcus aureus (MRSA). The study, published in Cell on 14 August, details how the team computationally generated more than 36 million hypothetical molecules and filtered them with machine-learning models trained to predict antibacterial activity, toxicity and novelty. The leading compound, dubbed NG1, cleared drug-resistant N. gonorrhoeae infections in a mouse model, while a second molecule, DN1, eradicated MRSA skin infections in vivo. Both candidates disrupt bacterial cell membranes through mechanisms unlike those of existing antibiotics, potentially making it harder for microbes to evolve resistance. The work is part of MIT’s Antibiotics-AI Project led by biomedical engineer James Collins. Non-profit Phare Bio is now refining NG1 and DN1 for further preclinical testing, a process researchers say could take one to two years before the compounds are ready for formal safety studies. Health authorities estimate that antimicrobial resistance contributes to about five million deaths a year worldwide. By venturing beyond existing chemical libraries, the researchers believe their AI-driven approach could open a pipeline of treatments for other hard-to-treat bacteria, including Mycobacterium tuberculosis and Pseudomonas aeruginosa, though human trials remain several years away.