Can AI outsmart superbugs? MIT’s new antibiotics offer hope for cure

Times News NetworkDiagnosing disease and prompting treatment options is one thing, but can artificial intelligence really design new drugs to tackle the scourge of antimicrobial resistance? Apparently it can, right down to the atom.Massachusetts Institute of Technology (MIT) has used generative AI to structure two new antibiotics that have shown much promise in mice models, killing superbugs that cause drug-resistant gonorrhoea and MRSA (Methicillin-resistant Staphylococcus aureus), a bacterial infection many antibiotics fail to cure. Antimicrobial resistance has become a global public health threat. In 2019, it led to around 5 million deaths, according to World Health Organisation estimates. Excessive use of antibiotics has made bacteria mutate to dodge the drugs, but very few new antibiotics have been launched over the decades. The MIT study, published in the journal Cell, aims to bridge that gap.How were the drugs made?First, the scientists used AI algorithms to come up with 36 million possible compounds (real and imagined), which were screened for their antimicrobial potential. “The top candidates they discovered are structurally distinct from any existing antibiotics, and appear to work by novel mechanisms that disrupt bacterial cell membranes. This approach allowed the researchers to generate and evaluate theoretical compounds never been seen before,” wrote Anne Trafton for MIT News.The researchers then trained the AI to identify how different molecular structures (of carbon, nitrogen, hydrogen, oxygen and other atoms) affect bacteria. Based on that, they tried two approaches to design two new antibiotics. The first got the AI to scan millions of chemical fragments to look for those that had the size of eight to 19 atoms, and build on that. The second approach let the AI design the drug freely. Both approaches filtered out potentially toxic compounds or anything that was too similar to existing antibiotics. The first strategy resulted in a new potential medicine for sexually transmitted gonorrhoea, and the second one targeted MRSA, which is a gram-positive bacteria that lives on skin but can lead to severe infection when it enters the body. The study’s senior author and MIT professor James Collins told BBC that it was exciting to see how generative AI can be used to design completely new antibiotics. “AI can enable us to come up with molecules cheaply and quickly and, in this way, expand our arsenal and really give us a leg up in the battle of our wits against the genes of superbugs,” he added.Long haul before actual useThough the drugs are a sure hit in the lab, it might take years before they can be used on humans. They need to be perfected for clinical trials first and that might prove a lengthy process. Dr Andrew Edwards from Imperial College, London called it significant research with huge potential but flagged the stumbling blocks as well. “While AI promises to dramatically improve drug discovery and development, we still need to do the hard yards when it comes to testing safety and efficacy,” he told BBC, adding that getting the experimental medicines ready could be a long and costly affair, with no surety of being prescribed to people in the end.Collins admitted that better AI models are needed for drug discovery so that they can effectively transition from the trial stage to actually benefit the human body. And Chris Dowson from the University of Warwick pointed out that even if they do, their efficacy would depend on restricted use, as too much of the medicine would renew drug resistance fears. “How do you make drugs that have no commercial value?” he told the BBC.