Scientist Denise Catacutan works on experimental antibiotic discoveries with the help of artificial intelligence. MCMASTER UNIVERSITY
With the help of artificial intelligence (AI), a group of scientists has discovered a new antibiotic that can kill a deadly species of superbug.
AI could be made to reduce a list of thousands of chemical compounds to just a few that have been developed in a laboratory.
The result was a powerful experimental antibiotic called abaucin, which must undergo further testing before it can be used in treatment.
Researchers in Canada and the United States say AI has the power to greatly accelerate the discovery of new drugs.
It’s the latest example of how powerful this tool can be in science and medicine.
Don’t be superbugs
Antibiotics kill bacteria. However, new ones are needed for decades, and bacteria are becoming more difficult to treat every day, since we already have resistance to antibiotics.
It is estimated that more than one million people a year die from infections caused by antibiotic-resistant treatment agents.
The researchers focused on one of the most troublesome species of bacteria: Acinetobacter baumannii, which can infect wounds and pneumonia.
Although you don’t hear it often, it’s one of three superbugs that the World Health Organization recognizes as a “critical” threat.
It is often able to bypass multiple antibiotics and cause problems in hospitals and nursing homes, where it can survive on surfaces and medical equipment.
Dr Jonathan Stokes, a researcher at McMaster University in Canada, describes this superbug as “public enemy number”, as it is “common” to find cases where “it is resistant to almost all antibiotics.”
Dr. Jonathan Stokes describes the superbug Acinetobacter baumannii as “Public Enemy Number.” MCMASTER UNIVERSITY
Artificial intelligence
To find a new antibiotic, researchers first used AI. They took thousands of drugs, whose precise chemical structure was known, and manually screened Acinetobacter oeni to identify which one could slow down or kill the superbug.
This information was fed into the AI to learn the chemical characteristics of the drugs that could target the bacteria.
The AI then reported a list of 6,680 compounds whose efficacy was unknown. The results, published in the scientific journal Nature Chemical Biology, show that it took the instrument an hour and a half to produce the short list.
The researchers tested 240 in the lab and found nine potential antibiotics. One of them was the powerful antibiotic abaucin.
Laboratory experiments have shown that wounds in infected mice can be treated and samples of A. baumannii can kill patients.
“This is when the work starts,” Stokes said.
The next step is to develop the drug in the laboratory and then conduct clinical trials.
Stokes warns that the first antibiotics discovered with the help of AI can be prescribed until 2030.
Interestingly, this experimental antibiotic had no effect on other bacteria species and only worked on A. baumannii.
Many antibiotics kill bacteria indiscriminately. The researchers believe that abaucin’s precision would make it more difficult for drug resistance to develop and could lead to fewer side effects.
Bacteria are grown in the laboratory. GETTY PICTURE
In principle, the AI could detect tens of thousands of potential compounds, which cannot be done manually.
“AI increases the rate and, in a perfect world, decreases the cost of finding these new classes of antibiotics that we desperately need,” Stokes noted.
Researchers have approved the use of AI in the discovery of antibiotics with E. coli bacteria in 2020, but now they are focusing on other infectious agents: Staphylococcus aureus and Pseudomonas aeruginosa.
“This discovery further supports the premise that AI can significantly accelerate and expand the search for new antibiotics,” said Professor James Collins at the Massachusetts Institute of Technology.
“This groundbreaking work will help show that we can use AI to combat problematic pathogens like A. baumannii,” he added.
