Skip to content Skip to footer

MIT researchers have utilized deep learning, a form of artificial intelligence, to discover a series of compounds capable of killing a drug-resistant bacterium responsible for over 10,000 deaths in the United States annually. The research, published in Nature, demonstrates these compounds can kill methicillin-resistant Staphylococcus aureus (MRSA) in lab dishes and two mouse models. Apart from their compelling potency, these compounds exhibit low toxicity against human cells, marking them as promising drug candidates.

A significant breakthrough in the study was the researchers’ ability to understand how the deep-learning model made its antibiotic potency predictions, potentially guiding the creation of even more effective drugs. This deep insight into the learning process of the model establishes a time and resource-efficient framework for advancing drug development.

The study is an outcome of MIT’s Antibiotics-AI project led by James Collins. With a mission to uncover new antibiotics against seven lethal bacteria over seven years, this project has already produced potential drugs against Acinetobacter baumannii and other drug-resistant bacteria.

The researchers trained the deep learning model with expansive datasets, testing around 39,000 compounds for antibiotic activity against MRSA. By utilizing an algorithm known as Monte Carlo tree search, they were able to understand how the model was making its predictions. The algorithm enabled the model to provide an estimate of each molecule’s antimicrobial activity and predict which substructures of the molecule likely account for that activity.

To refine the pool of candidate drugs, the researchers trained three additional deep learning models to predict whether the compounds would be toxic to three different types of human cells. Combining this data with the predictions of antimicrobial activity, they discovered compounds that could kill microbes while having minimal adverse effects on humans. After screening roughly 12 million commercially available compounds, the models identified five different classes of compounds predicted to be active against MRSA. After testing 280 compounds, two proved to be promising antibiotic candidates.

The compounds appear to kill bacteria by interrupting their ability to maintain an electrochemical gradient across their cell membranes, essential for numerous critical cell functions. Collins’ lab had discovered a similar method in 2020 with the antibiotic candidate Halicin, but it was specific to Gram-negative bacteria, while MRSA is a Gram-positive bacterium.

The research team has shared their findings with Phare Bio, a nonprofit started by Collins and others as part of the Antibiotics-AI Project, which is now planning a comprehensive analysis of the chemical properties and potential clinical use of these compounds. Meanwhile, Collins’ lab is extrapolating their findings to design additional drug candidates and seek compounds that can kill other types of bacteria.

Leave a comment

0.0/5