Scientists use AI to discover new antibiotic

Scientists from McMaster University and the Massachusetts Institute of Technology (MIT) have leveraged artificial intelligence (AI) to uncover a new antibiotic that could be used to combat a lethal drug-resistant pathogen, Acinetobacter baumannii, that is primarily seen in hospital associated infections.  

 The researchers used an AI algorithm to predict new structural categories of antibacterial compounds. This process allowed researchers to access a vast number of molecules with antibacterial properties and to successfully identify a new antibacterial compound, named abaucin.  

The World Health Organization (WHO) classified A. baumannii, as a “priority pathogen” that poses one of the greatest risks to human health, because of its ability to survive on surfaces for extended periods and for its ability to acquire genes associated with antibiotic resistance from other bacterial species.  

A. baumannii can lead to life-threatening conditions such as pneumonia, meningitis, and wound infections. Once in a hospital, it is difficult to eradicate from both the patient and environment.

Researchers used a specially trained AI algorithm to screen through thousands of potential compounds with antibacterial properties and predict new chemical structures as solutions.  

“We had a whole bunch of data that was just telling us about which chemicals were able to kill a bunch of bacteria and which ones weren’t. My job was to train this model, and all that this model was going to be doing is telling us essentially if new molecules will have antibacterial properties or not,” said Gary Liu, a graduate student at MacMaster University and one of the contributing researchers on the study. 

 “Then basically through that, we’re able to just increase the efficiency of the drug discovery pipeline and … hone in [on] all the molecules that we really care about,” he added.  

After the AI was trained, the scientists utilized it to analyze a set of 6,680 compounds that were previously unfamiliar to the model. The analysis took about an hour and a half and provided researchers with several hundred compounds as potential antibiotic solutions. Of those compounds, 240 underwent rigorous laboratory testing, resulting in the identification of nine promising antibiotics, including abaucin. 

Abaucin successfully suppressed an A. baumannii wound infection in mice, but it could take years before the drug comes to market.  

“Using AI, we can rapidly explore vast regions of chemical space, significantly increasing the chances of discovering fundamentally new antibacterial molecules,” said Jonathan Stokes, lead author on the study and an assistant professor in McMaster’s Department of Biomedicine & Biochemistry. 

The methods they used could also help expedite the discovery of more antibiotics against drug-resistant bacterial strains.  

“We know broad-spectrum antibiotics are suboptimal and that pathogens have the ability to evolve and adjust to every trick we throw at them,” Stokes shared. “AI methods afford us the opportunity to vastly increase the rate at which we discover new antibiotics, and we can do it at a reduced cost. This is an important avenue of exploration for new antibiotic drugs.” 

The researchers’ discovery is the beginning of what looks like a promising road in drug development.