Every 15 minutes someone in the United States dies from a drug-resistant infection (Center for Disease Control and Prevention, 2019). Antimicrobial resistance occurs when bacteria, viruses, fungi, and parasites develop defenses against the drugs designed to destroy them. These antimicrobial-resistant germs are called “superbugs” – a villainous title that is rather fitting. Superbugs threaten the amazing health benefits of antibiotics. The World Health Organization has declared antimicrobial resistance one of the top ten threats for global health (World Health Organization, 2019), leading to lengthier hospital stays, increased chance of fatality, and a higher risk of disease transmission (McIntosh, 2018). There are 4.5 million annual deaths associated with antimicrobial resistance globally (World Health Organization, 2023). By 2050, the United Nations estimates that over 10 million deaths per year will be directly attributed to antimicrobial resistance (UN Environment).
A particularly dangerous superbug that has emerged in recent decades is MRSA (methicillin-resistant Staphylococcus aureus). MRSA is particularly dangerous because of its ability to resist an entire class of drugs called beta-lactams. These include commonly prescribed drugs like penicillin, amoxicillin, and oxacillin (Baylor College of Medicine).
To combat these difficulties in antibiotic resistance, a team of researchers from MIT is using deep learning, a sect of machine learning, to discover compounds that can kill MRSA (Wong et.al, 2023). Deep learning is a subset of machine learning that uses neural networks to recognize patterns in large datasets. By analyzing these patterns, it learns to make predictions or classifications on unseen data. Researchers have shown that deep learning models for drug discovery can be applied to more tasks beyond just identifying specific targets that drug molecules bind to. Felix Wong at the Broad Institute of MIT and Harvard in Massachusetts explains, “our [machine learning] models tell us not only which compounds have selective antibiotic activity, but also why, in terms of their chemical structure” (Hsu, 2023).
To accomplish this feat, researchers screened the antibiotic activities and the potential toxicity to human cells, otherwise known as cytotoxicity, of a whooping 39,312 compounds. The results were then used as training data for the deep learning models so that they could recognize the patterns in the chemical atoms and bonds of each compound. Using this trained ensemble of models, researchers then predicted the antibiotic activity and cytotoxicity for 12,076,365 compounds with potential for killing MRSA. Afterwards, the researchers estimated each molecule’s antimicrobial activity and identified which substructures are likely responsible for that activity. They empirically tested 283 compounds with high predicted antibiotic activity and low predicted cytotoxicity. One of these classes of structural compounds was shown to be resistant against MRSA. The compounds were evaluated in two mouse models: one for MRSA skin infection and another for MRSA systemic infection. Each compound disrupted the electrochemical gradient across bacterial cell membranes, reducing the MRSA population by a factor of 10 (Wong et.al, 2023).
This approach offers a powerful strategy for overcoming antimicrobial resistance and treating serious infections, marking a major step forward in combating resistant infections. James Collins, the Termeer Professor of Medical Engineering and Science at MIT says “it is time-efficient, resource-efficient, and mechanistically insightful, from a chemical-structure standpoint, in ways that we haven’t had to date.” (Trafton, 2023). The researchers’ breakthrough represents a significant advancement in the fight against drug-resistant bacteria and underscores the promise of innovative methodologies in modern medicine.
References
Baylor College of Medicine. (n.d.). Methicillin-resistant Staphylococcus aureus (MRSA). https://www.bcm.edu/departments/molecular-virology-and-microbiology/emerging-infections-and-biodefense/specific-agents/mrsa
Centers for Disease Control and Prevention. (2019). 2019 Antibiotic Resistance Threats Report. Antimicrobial Resistance. https://www.cdc.gov/antimicrobial-resistance/data-research/threats/index.html
Collins lab. The Audacious Project. (n.d.). https://www.audaciousproject.org/grantees/collins-lab
Hsu, J. (2023, December 20). Ai discovers new class of antibiotics to kill drug-resistant bacteria. New Scientist. https://www.newscientist.com/article/2409706-ai-discovers-new-class-of-antibiotics-to-kill-drug-resistant-bacteria/
McIntosh, J. (2018). Antimicrobial and antibiotic drug resistance: Causes and more. Medical News Today. https://www.medicalnewstoday.com/articles/283963#causes
Trafton, A. (2023, December 20). Using AI, MIT researchers identify a new class of antibiotic candidates. MIT News | Massachusetts Institute of Technology. https://news.mit.edu/2023/using-ai-mit-researchers-identify-antibiotic-candidates-1220
UN Environment. (n.d.). Antimicrobial resistance: A global threat. UNEP. https://www.unep.org/topics/chemicals-and-pollution-action/pollution-and-health/antimicrobial-resistance-global-threat
Wong, F., Zheng, E. J., Valeri, J. A., Donghia, N. M., Anahtar, M. N., Omori, S., Li, A., Cubillos-Ruiz1, A., Krishnan, A., Jin, W., Manson, A. L., Friedrich, J., Helbig, R., Hajian, B., Fiejtek, D. K., Wagner, F. F., Soutter, H. H., Earl, A. M., Stokes, J. M., … Collins, J. J. (2023, December 20). Discovery of a structural class of antibiotics with explainable deep learning. https://dspace.mit.edu/handle/1721.1/153216
World Health Organization. (2023, November 21). Antimicrobial resistance. https://www.who.int/news-room/fact-sheets/detail/antimicrobial-resistance#:~:text=It%20is%20estimated%20that%20bacterial,development%20of%20drug%2Dresistant%20pathogens
World Health Organization. (2019). Ten health issues who will tackle this year. World Health Organization. https://www.who.int/news-room/spotlight/ten-threats-to-global-health-in-2019