AI Spots a New Antibiotic Effective Against Deadly, Drug‑Resistant Bacteria
An artificially intelligent program has identified a new antibiotic that is effective against many drug-resistant strains of bacteria.
India is the antibiotic-resistant capital of the world. In 2018, 130,000 people were diagnosed with multi-drug resistant tuberculosis, according to Stop TB, which adds the number is likely closer to 200,000, when underreporting is considered.
The new antibiotic also appears effective, in tests on mice, against drug-resistant strains of Acinetobacter baumannii and Enterobacteriaceae, two of the three most dangerously antibiotic-resistant bacteria that pose the “greatest threat to human health,” according to the World Health Organization.
Antibiotic resistance develops through the overuse of antibiotics in humans and livestock and antibiotic pollution of waterways by pharma companies. Between 2010 and 2015, antibiotic use doubled in India thanks to greater access and higher incomes, not necessarily because it was needed.
The result? “Diseases like pneumonia and typhoid have become difficult to treat. In 70% of the cases, treatment begins with more expensive, third-generation drugs, which are administered for a longer duration than before,” Sunil Gupta, additional director of the Delhi-based National Centre for Disease Control (NCDC), told LiveMint.
Once a world-changing, life-saving treatment after its early-20th century discovery, antibiotics are increasingly becoming obsolete – with no substitute therapy in sight. Last year, the World Health Organization issued a warning that antibiotic-resistant infections could cause 10 million deaths each year and cause a loss of US$ 1 trillion globally by 2050, if no action is taken or advances made.
“People keep finding the same molecules over and over,” synthetic biologist Jim Collins at the Massachusetts Institute of Technology, told the journal Nature. “We need novel chemistries with novel mechanisms of action.”
Here comes AI to save the day.
Related on The Swaddle:
Led by Collins, researchers at the Massachusetts Institute of Technology (MIT) built an algorithm that could comb through a limited library of 2,335 antibacterial molecules and trained it to identify the ones capable of fighting the common bacteria E.coli. With that knowledge instilled, they then set the algorithm to trawling a 6,000-strong collection of molecular compounds already used or in testing to treat a variety of diseases, not just bacterial infections. The algorithm was also programmed to screen out results that resemble antibiotics known and in use already.
Published in the journal Cell, the researchers reported that among several promising candidates, the algorithm identified a compound originally developed to treat diabetes, but which never made it to market. The researchers dubbed it halicin, after HAL, the AI character from 2001: A Space Odyssey. They then tested halicin and several other compounds on a variety of infections in mice, including Mycobacterium tuberculosis, C.difficile, drug-resistant Enterobacteriaceae, and multidrug-resistant Acinetobacter baumannii. Halicin eliminated each infection.
The researchers then set the algorithm onto a library of 1.5 billion compounds, from which the AI identified 23 with antibiotic potential. Tests on mice revealed two of them to be especially effective.
Experts in the field are heralding the discovery as much for the hope offered by the identified compounds as for the quickness of discovery, which going forward could revolutionize research into new drugs not only to fight bacteria, but also to fight cancers and other maladies.
“I think it’s a breakthrough in a field of much unmet need,” César de la Fuente, a bioengineer at the University of Pennsylvania who works on AI and antibiotics, told Vox. de la Fuente was not involved in the MIT study. “After all, no new classes of antibiotics have been discovered for decades. This one is definitely structurally different from conventional antibiotics.”
That said, experts are also sounding a note of caution, too: as promising as the AI’s findings are, many compounds work in mice but aren’t effective in humans. Even if the ones discovered by the algorithm prove effective on humans, too, and (after much testing and many years) become available on the market, they don’t really solve the underlying problem — antibiotic overuse. Instead, they only buy us a little more time.