From Artificial Intelligence a new antibiotic in just 2 hours

A powerful new antibiotic against the superbug Acinetobacter baumannii, identified by the World Health Organization as one of the most dangerous resistant bacteria in the world, has been discovered thanks to Artificial Intelligence, through a process that promises to revolutionize this field: the new molecule, in fact, was found by the algorithm among almost 7,000 candidates in just 2 hours. The result, published in the journal Nature Chemical Biology by a group led by Canada's McMaster University, paves the way for the possibility of greatly speeding up the process of discovering new antibiotics for many other microorganisms that threaten human health and beyond.

The discovery of new antibiotics against A. baumannii, which is often found in hospitals and can lead to serious infections, through traditional methods has proved to be a challenge, due to the large amount of time, high costs and poor results. Researchers led by Gary Liu, Denise Catacutan and Khushi Rathod then turned to an algorithm capable of accessing hundreds of millions, perhaps billions, of molecules with antibacterial properties.

To obtain the data with which to train it, the authors of the study first cultivated the bacterium in the laboratory, exposing it to about 7,500 different chemical compounds: the structure of each of these was then fed to the algorithm, along with the information associated with their effectiveness. Once ready, the AI sifted through a library of 6,680 never-before-evaluated molecules, providing the 240 most promising candidates in a couple of hours.

At this point, the ball is in the researchers' court, who by testing the compounds in the laboratory have identified the new antibiotic: its peculiarity lies in the fact that it targets only A. baumannii, a characteristic that prevents the bacterium from rapidly developing resistance to the drug.