AI Develops New Antibiotics for Drug-Resistant Infections
Researchers have announced that artificial intelligence has crafted two promising antibiotics potentially effective against drug-resistant strains of gonorrhoea and MRSA. These drugs were meticulously designed atom-by-atom by the AI and have shown the ability to eliminate these superbugs in laboratory and animal trials.
Although these compounds still require significant refinement and several years of clinical testing before they can be prescribed to patients, the Massachusetts Institute of Technology (MIT) team behind the study believes that AI could usher in a “second golden age” for antibiotic development.
Antibiotics are meant to eradicate bacteria, yet infections that resist these treatments now result in more than a million deaths annually. The overuse of antibiotics has enabled bacteria to evolve and evade treatment, leading to a long-standing shortage of new antibiotics.
Previously, researchers have employed AI to sift through countless chemicals to spot candidates that could be developed into new antibiotics. In this instance, the MIT team took a leap forward by using generative AI to create antibiotics specifically aimed at treating gonorrhoea and MRSA.
Their research, detailed in the journal Cell, explored a staggering 36 million compounds, including those not yet discovered. Scientists trained the AI model by providing it with the chemical structures of known antibiotics and corresponding data on their efficacy against various bacterial species. This allowed the AI to learn how different molecular designs, composed of elements like carbon, hydrogen, oxygen, and nitrogen, affect bacterial growth.
Two distinct strategies were employed in the AI’s design of these antibiotics. One approach identified a valuable starting point by searching through a vast library of chemical fragments, ranging from eight to nineteen atoms long, before building upon that. The other approach gave the AI complete freedom to innovate right from the start.
The design process aimed to avoid anything resembling existing antibiotics and to ensure that the research focused on creating viable medicines rather than non-therapeutic substances. There was also a keen emphasis on filtering out compounds predicted to be toxic to humans.
After the promising designs were generated, they underwent testing against bacteria in controlled lab settings and on infected mice, leading to the creation of two potential new drugs.
“We’re thrilled because this demonstrates that generative AI can effectively design new antibiotics,” said Prof. James Collins from MIT. “AI can help us develop molecules quickly and cost-effectively, expanding the arsenal against superbugs.” However, he noted that the drugs still require refinement—a process expected to take one to two years—before clinical trials can commence.
Dr. Andrew Edwards from the Fleming Initiative and Imperial College London characterized the work as “very significant,” emphasizing its “enormous potential” and unique approach to identifying new antibiotics. Nevertheless, he cautioned that despite the promise of AI, thorough testing for safety and efficacy remains essential.
This testing phase could be lengthy and costly, and there’s no guarantee that the resulting medicines will eventually be approved for use. Some experts urge a more comprehensive application of AI in drug discovery, with Prof. Collins advocating for improved models that predict drug performance in actual biological systems, not just in laboratory settings.
Manufacturing also poses challenges for these AI-designed drugs. Of the top 80 theoretical treatments for gonorrhoea, only two have been successfully synthesized into actual medicines. Prof. Chris Dowson from the University of Warwick expressed that the study represents a considerable advancement in employing AI for antibiotic discovery, potentially addressing the pressing issue of bacterial resistance.
However, there are economic dilemmas involved. When it comes to drug-resistant infections, the question arises: how to create profitable drugs that, ideally, should be used sparingly to maintain their effectiveness? It’s a complex problem that needs addressing as new antibiotics emerge.
