Google DeepMind has published a paper in Nature describing an AI system that identified 11 novel antibiotic compounds effective against drug-resistant bacteria. The system screened over 200 million molecular structures using a graph neural network trained on known antimicrobial activity data.
Three of the discovered compounds showed exceptional potency against methicillin-resistant Staphylococcus aureus in laboratory tests, with minimal toxicity to human cells. The researchers are collaborating with pharmaceutical partners to advance two candidates into preclinical animal studies later this year.
The breakthrough addresses a critical public health concern, as the World Health Organization estimates that antimicrobial resistance could cause 10 million deaths annually by 2050 without the development of new therapeutic options.
