Deep learning AI discovers surprising new antibiotics

By | February 25, 2020

With the rapid rise in drug resistance in many pathogens, new antibiotics are desperately needed. Yet few new antibiotics have entered the market of late, and even these are just minor variants of old antibiotics. While the prospects look bleak, a recent article by Sriram Chandrasekaran, assistant professor of biomedical engineering, explores how researchers hunting for new drugs are embracing artificial intelligence (AI).

Drug researchers using a new type of deep learning AI called graph neural networks have found a gold mine of new candidates that do not look anything like existing antibiotics. “Unlike people, AI has no preconceived notions, especially about what an antibiotic should look like,” writes Chandrasekaran. “Using old-school AI, my lab recently discovered some surprising candidates for treating tuberculosis, including an anti-psychotic drug.”