Regina Barzilay, faculty lead for artificial intelligence (AI) at the MIT Jameel Clinic, was amongst the team of MIT researchers that have found two new antibiotic compounds since 2019. The discovery of new antibiotics, in this case Halicin and Abaucin, is a rarity in modern medicine, however the method used to identify them is becoming increasingly common across the scientific community.
Researchers used an AI model that had been trained on the chemical structures and efficacy of several thousand known antibiotics. The model identified links between chemical structures and their ability to damage bacteria, and produced a list of the most promising structures, which researchers lab-tested and identified the antibiotics. The AI-assisted process saved researchers time and money in the trial-and-error phase. Regina compares drug discovery to finding a needle in a haystack, adding that AI acts as a metal detector.
AI is now assisting researchers across a range of fields, often performing tasks that were previously painstaking and expensive. AI is more adept than supercomputers at managing the constant inflow of data required for weather forecasting, for example. Generative adversarial networks are being applied to sperm-whale vocalisations as researchers make progress in understanding whale communication. The Large Hadron Collider generates more data in an hour than Facebook does in a year. Scientists used machine learning to filter through the data, which led to the discovery of the Higgs boson.
As AI models continue to improve and their applications expand, scientific progress could happen more quickly and at less cost. Speaking on where AI could take us, Regina says: "The type of questions that we will be asking will be very different from what we’re asking today."