In the shadow of the Covid-19 pandemic, super bacteria pose a growing health threat. New antibiotics are needed to reverse the trend, but their development takes time. This IBM-designed AI promises to speed things up.
A major asset in the fight against antibiotic resistance
While the discovery of penicillin represented one of the most important scientific breakthroughs of the 20th century, making it possible to treat previously fatal infections, decades later, the situation is tending to be reversed. Like all organisms, bacteria adapt to their environment and to the drugs we take. In the absence of new antibiotics or other treatments, scientists predict that infections once considered minor could claim up to 10 million lives per year by 2050.
More worryingly, the development of new compounds takes years and involves an enormous amount of trial and error, due to the myriad chemical combinations possible for potential molecules. To remedy this and reduce their development time, researchers are increasingly relying on artificial intelligence.
Recently described in the review Nature, the systemAI created by scientistsIBM Research allows you to explore all the possibilities of molecular configurations in record time.
Initially, a model called ” deep generative autoencoder Examines a series of peptide sequences, highlighting essential information about the functions and molecules that compose them and looking for similarities with other peptides. Once this step is completed, the system CLaSS (Controlled Latent attribute Space Sampling) takes over and relies on the data collected to generate new peptide molecules with specific and desired properties (antimicrobial efficacy in this specific case).
20 candidate antibiotic peptides synthesized and tested in 48 days
Of course, the ability to kill bacteria is not the only requirement for an antibiotic. This should also be safe for human use and ideally work on a range of classes of bacteria. The molecules generated byAI are therefore subjected to deep learning classifiers to eliminate ineffective or toxic combinations.
In 48 days, the systemAI identified, synthesized and tested 20 novel candidate antibiotic peptides. Two of them turned out to be particularly promising. Very potent against a series of bacteria from the two main classes (Gram positive and Gram negative), these punctured holes in the outer membranes of the latter and also exhibited low toxicity and seemed very unlikely to cause new drug resistance in E. coli, in tests carried out on rodents and cell cultures.
As interesting as the two candidate antibiotics designed, the real breakthrough obviously remains the approach used. Being able to quickly develop and test new compounds could help avoid the nightmarish scenario of a return to a world without antibiotics.