Melis Anahtar remembers the frustration in her clinic at Massachusetts General Hospital when another patient tested positive for a strain of gonorrhea that no existing antibiotic could reliably treat. With over 600,000 reported cases in the U.S. each year and rising resistance to every major drug class, the search for new treatments has felt like a race against time. Now, in a breakthrough led by Anahtar and James Collins at the Wyss Institute at Harvard and MIT, artificial intelligence has scanned 6 million chemical compounds and surfaced two promising new candidates that could become the next generation of antibiotics for drug-resistant gonorrhea. This isn’t just a step forward—it’s a new playbook for fighting superbugs.
Gonorrhea, caused by the bacterium Neisseria gonorrhoeae, is a master of adaptation. It has developed resistance to sulfonamides, penicillins, tetracyclines, fluoroquinolones, and most recently, cephalosporins. The recent approval of zoliflodacin and gepotidacin—two new oral antibiotics—offered hope, but as Anahtar warns, “If these two antibiotics get used broadly, it's nearly guaranteed that the pathogen will develop significant resistance to them eventually.” The cycle repeats: a new drug emerges, resistance follows within five to ten years, and the pipeline runs dry. To break this pattern, scientists needed a faster, smarter way to discover drugs.
Enter AI. The team first tested 38,650 small molecules for activity against N. gonorrhoeae, using the results to train a deep learning model capable of recognizing chemical structures with antimicrobial potential. Unlike traditional methods that rely on known antibiotic frameworks, this model was designed to find entirely novel compounds—molecules that might target obscure bacterial pathways and thus evade existing resistance mechanisms. Once validated, the AI was unleashed on a library of six million compounds. From that vast chemical universe, it identified 213 promising candidates. After rigorous lab testing, two stood out: they not only killed multidrug-resistant strains but did so with high selectivity, meaning they spared human cells and showed low potential for driving resistance.
The implications extend far beyond gonorrhea. This study, published in Science Translational Medicine, demonstrates that AI can accelerate antibiotic discovery from years to months, opening a new frontier in the fight against antimicrobial resistance. These two lead compounds are now being optimized for clinical development, bringing hope that the medical community can stay one step ahead of evolving pathogens.
As Collins puts it, “We’ve arrived at an incredibly important point in time” — where machine learning meets medicine, and where the next life-saving drug might be hidden in a dataset, waiting to be found.