MIT Researchers Use AI To Discover Possible Superbug-Killing Antibiotics

By Toby T

Exploits being made by artificial intelligence in the healthcare space have been especially impressive. Recently, a new AI-based discovery has had many scientists and healthcare experts talking. 

Source: @roberto_sorin/Unsplash

Researchers from MIT recently used a specialized computer algorithm to parse a massive digital archive containing over 100 million chemical compounds in order to find those that could kill bacteria using different formulations and mechanisms from existing medications. 

According to a report from 2020, this method especially highlighted a molecule that seemed to possess some fascinating antibiotic properties. And the resulting molecule was named halicin; a reference to “Hal,” a sentient AI system from the popular movie 2001: A Space Odyssey.

In trials involving mice, halicin was able to properly treat tuberculosis and Enterobacteriaceae. Just as well, the formulation proved to be especially effective against Clostridium difficile, a “stomach bug” that is especially prominent in hospitals. 

As the scientists explained, their discovery could easily be one of the most powerful antibiotics ever formulated. Interestingly, the molecule has a distinct structural makeup – one which would have made it especially difficult to find without the use of AI.  

Source: @cdc/Unsplash

Following the discovery of halicin, the researchers went back to the database and used the same AI algorithm to find other possible candidates. And, in just three days, 23 candidates were discovered; each possessing different molecular structures than existing antibiotics. 

Further tests showed that 8 of these 23 candidates had antibacterial properties – with a couple of them being especially potent. As the scientists explain, all these candidates could end up being important to the tracking and treatment of superbugs and other antibiotic-resistant infections. 

According to estimates, no less than 2.8 million people get antibiotic-resistant infections annually in the United States alone, with 350,000 deaths. Nevertheless, through the assistance of AI, we can anticipate a significant reduction in this figure.