Artificial intelligence has revolutionized the medical world, designing entirely new antibiotics to combat drug-resistant bacteria. Using generative AI, scientists at MIT have successfully developed powerful molecules that are structurally different from known antibiotics. These new-generation antibiotics could play a significant role in combating resistant infections, particularly gonorrhea and hospital-acquired bacterial infections (MRSA).
AI-powered antibiotic design
The research team used two different AI methods for this discovery. The first approach, “fragment-based design,” scanned a massive database of 45 million chemical particles. The AI identified particles with potential antibacterial activity and eliminated those that were toxic to humans or similar to existing antibiotics, leaving approximately 1 million candidates.
A selected fragment from these candidates was processed using AI algorithms (CReM and F-VAE) to create millions of new molecules. One of these, “NG1,” effectively eliminated drug-resistant gonorrhea bacteria (Neisseria gonorrhoeae) in laboratory and mouse tests.
In the second method, “free design,” no restrictions were placed on the AI. CReM and VAE algorithms generated over 29 million theoretical molecules. Of these, 90 were selected, 22 were produced, and six were found to exhibit potent antibacterial activity against MRSA. Of these molecules, “DN1” was successful in treating skin infections in mice.
The most striking feature of these developed molecules is their structure, which is completely different from known antibiotics. According to the research, NG1 targets the LptA protein in the bacterial outer membrane, disrupting membrane synthesis and killing the cell.
DN1, on the other hand, works by disrupting membrane integrity through a broader mechanism of action. This demonstrates the potential of AI to develop not only derivatives of existing drugs but also drugs with entirely new mechanisms of action.
MIT Professor of Biomedical Engineering James Collins explained the significance of this discovery: “We’re using AI not just to screen existing libraries, but to generate new molecules in unexplored chemical spaces. This opens up a previously unexplored avenue of discovery for antibiotic development.”
This work could usher in a new era in the fight against bacteria that cause more than 5 million deaths each year and are becoming increasingly resistant to existing antibiotics. Studies on NG1 and DN1 are ongoing, pre-clinical trials. The team plans to use the same AI platform against other dangerous bacteria, such as tuberculosis and pseudomonas.
