Scientists have used artificial intelligence (AI) to discover a new antibiotic that can kill a deadly species of superbug
Harnessing the capabilities of AI, researchers efficiently sifted through numerous chemical compounds to identify a select few for subsequent laboratory experimentation. This meticulous process led to the development of abaucin, an experimental antibiotic exhibiting promising potency against deadly superbugs. However, additional testing is required before its practical application can be determined.
According to scientists from Canada and the United States, the utilization of AI has the potential to greatly expedite the process of discovering new drugs. This breakthrough serves as a prime illustration of how artificial intelligence tools can wield a revolutionary impact on the fields of science and medicine.
Stopping the superbugs
The effectiveness of antibiotics in eliminating bacteria has been widely recognized. Nevertheless, there has been a scarcity of new antibiotic drugs over the past few decades, resulting in bacteria developing resistance to existing treatments. As a consequence, the number of fatalities attributed to infections that are resistant to antibiotics is estimated to exceed one million individuals annually.
The researchers dedicated their efforts to addressing Acinetobacter baumannii, a highly troublesome bacterial species known for causing pneumonia and infecting wounds. Although it may not be widely recognized, it is classified as one of the three superbugs by the World Health Organization, indicating a “critical” threat to global health.
Frequently, Acinetobacter baumannii exhibits resilience against numerous antibiotics, making it a significant concern in hospital and care home settings. Its ability to persist on surfaces and medical equipment further exacerbates the problem. Dr. Jonathan Stokes of McMaster University characterizes this bacterium as “public enemy number one,” emphasizing its prevalence and the alarming frequency with which it demonstrates resistance to nearly all available antibiotics.
The researchers embarked on training the AI by utilizing thousands of drugs with well-defined chemical structures. They conducted manual tests on Acinetobacter baumannii to identify drugs that exhibited the ability to inhibit or eradicate the bacterium. The data obtained from these experiments were then fed into the AI system, enabling it to learn the specific chemical characteristics necessary to target the problematic bacteria effectively.
The AI was subsequently deployed to analyze a list of 6,680 compounds with unknown effectiveness. The findings, published in Nature Chemical Biology, revealed that the AI was able to generate a shortlist within just an hour and a half. From this shortlist, the researchers selected 240 compounds for laboratory testing, ultimately identifying nine potential antibiotics. Among these promising candidates was the highly potent antibiotic abaucin.
In laboratory experiments, abaucin demonstrated efficacy in treating infected wounds in mice and effectively eradicated A. baumannii samples obtained from patients. However, Dr. Stokes emphasized that this is just the beginning of the process. The next phase involves refining the drug in the laboratory and conducting clinical trials. Dr. Stokes anticipates that it may take until 2030 for the first AI-derived antibiotics to be ready for prescription.
Interestingly, the experimental antibiotic abaucin exhibited no impact on other bacterial species, specifically targeting A. baumannii. Unlike many antibiotics that have a broad-spectrum effect, the researchers believe that abaucin’s precision could make it more challenging for drug resistance to develop and potentially result in fewer side effects. The AI’s capability to screen tens of millions of potential compounds is a significant advantage, as manual screening of such a large number would be impractical.
According to Dr. Stokes:
“AI accelerates the pace and, ideally, reduces the cost of discovering these much-needed new classes of antibiotics.”
The researchers initially tested the application of AI in antibiotic discovery with E. coli in 2020 and have now applied their knowledge to tackle more challenging bacteria. Their future focus will be on examining Staphylococcus aureus and Pseudomonas aeruginosa.
Prof. James Collins from the Massachusetts Institute of Technology commented:
“This discovery provides additional evidence that AI can greatly speed up and broaden our quest for new antibiotics.”
He expressed excitement over the findings, stating:
“I’m thrilled to see that AI can assist in combating challenging pathogens like A. baumannii.”
Speaking on Radio 4’s The World Tonight, Prof Dame Sally Davies, the former chief medical officer for England and government envoy on antimicrobial resistance, expressed her optimism about the use of AI in antibiotic discovery. She stated, “We’re onto a winner,” and described the idea as a “big game-changer.” Prof Dame Sally Davies further expressed her excitement about the work of Dr Stokes, emphasizing that it has the potential to save lives.