In an effort to combat the rising issue of antibiotic resistance, which is associated with nearly 5 million deaths globally each year, researchers at Stanford Medicine and McMaster University have developed an artificial intelligence model, SyntheMol, capable of designing new potential drugs.
The AI was used to create structures and synthesis methods for six new drugs capable of battling resistant strains of Acinetobacter baumannii, a major cause of antibacterial resistance-related fatalities.
"Development of new antibiotics is a public health necessity," stated James Zou, an associate professor of biomedical data science at Stanford. Using artificial intelligence, scientists can generate molecules that haven't yet been tested or created in the natural world, potentially opening up new routes for effective drug use.
Prior to generative AI such as SyntheMol, researchers used computational methods to seek potential antibiotics, using algorithms to search through existing drug compounds to identify possibilities. Despite sifting through 100 million known compounds, this approach barely scratches the surface of potential drug-like molecules, estimated to be around 10^60.
The advantage of generative AI is its ability to "hallucinate", formulating entirely new responses. However, previous tests of this method resulted in impossible-to-synthesize molecules. To combat this, the researchers put restrictions on SyntheMol to ensure that the generated compounds could be created in a lab.
SyntheMol was trained on a library of more than 130,000 molecular building blocks and a range of chemical reactions, equipping the model with both the end results and the steps needed to reach them. SyntheMol produced around 25,000 potential antibiotics in under nine hours.
Researchers narrowed these suggestions down to the 70 compounds with the highest levels of potential efficacy. These were synthesized with the assistance of Ukrainian chemical company Enamine. Of these, six were able to kill a resistant strain of A. baumannii.