Using a type of artificial intelligence known as deep learning, MIT researchers have discovered a class of compounds that can kill a drug-resistant bacteria that causes more than 10,000 deaths in the United States each year.
in a study that appears today in NatureThe researchers showed that these compounds could kill methicillin-resistant people. Staphylococcus aureus (MRSA) cultured in a laboratory dish and in two mouse models of MRSA infection. The compounds also show very low toxicity against human cells, making them especially good drug candidates.
A key innovation of the new study is that the researchers were also able to discover what type of information the deep learning model was using to make its antibiotic potency predictions. This knowledge could help researchers design additional drugs that could work even better than those identified by the model.
“The idea here was that we could see what the models were learning to make their predictions that certain molecules would be good antibiotics. Our work provides a framework that saves time and resources and is mechanistically revealing, from a chemical structure point of view, in ways we have not had to date,” says James Collins, Termeer Professor of Engineering and Medical Sciences. at the Institute of Medical Engineering and Sciences (IMES) and the Department of Biological Engineering at MIT.
Felix Wong, a postdoc at IMES and the Broad Institute of MIT and Harvard, and Erica Zheng, a former graduate student at Harvard Medical School who was mentored by Collins, are lead authors of the study, which is part of the Antibiotics-ai Project at MIT. The mission of this project, led by Collins, is to discover new classes of antibiotics against seven types of deadly bacteria, over seven years.
Explainable predictions
MRSA, which infects more than 80,000 people in the United States each year, often causes skin infections or pneumonia. Severe cases can lead to sepsis, a life-threatening bloodstream infection.
In recent years, Collins and his colleagues at MIT's Abdul Latif Jameel Clinic for Machine Learning in Health (Jameel Clinic) have begun using deep learning to try to find new antibiotics. Their work has produced potential drugs against Acinetobacter baumanniia bacteria often found in hospitals, and many other drug-resistant bacteria.
These compounds were identified using deep learning models that can learn to identify chemical structures associated with antimicrobial activity. These models then screen millions of other compounds, generating predictions about which ones may have strong antimicrobial activity.
These types of searches have proven fruitful, but a limitation of this approach is that the models are “black boxes,” meaning there is no way to know what characteristics the model based its predictions on. If scientists knew how the models made their predictions, it would be easier for them to identify or design additional antibiotics.
“What we set out to do in this study was open the black box,” Wong says. “These models consist of a lot of calculations that mimic neural connections, and no one really knows what's going on under the hood.”
First, the researchers trained a deep learning model using substantially expanded data sets. They generated this training data by testing around 39,000 compounds for antibiotic activity against MRSA and then fed this data, as well as information about the compounds' chemical structures, into the model.
“You can represent basically any molecule as a chemical structure and also tell the model whether that chemical structure is antibacterial or not,” Wong says. “The model is trained with many examples like this. “If you then give it any new molecule, a new arrangement of atoms and bonds, it can tell you how likely that compound is predicted to be antibacterial.”
To figure out how the model made its predictions, the researchers adapted an algorithm known as Monte Carlo tree search, which has been used to help make other deep learning models, such as AlphaGo, more explainable. This search algorithm allows the model to generate not only an estimate of the antimicrobial activity of each molecule, but also a prediction of which substructures of the molecule are likely to represent that activity.
Powerful activity
To further narrow the pool of drug candidates, the researchers trained three additional deep learning models to predict whether the compounds were toxic to three different types of human cells. By combining this information with predictions of antimicrobial activity, the researchers discovered compounds that could kill microbes while having minimal adverse effects on the human body.
Using this collection of models, the researchers screened about 12 million compounds, all of which are commercially available. From this collection, the models identified compounds from five different classes, based on chemical substructures within the molecules, that were predicted to be active against MRSA.
The researchers purchased about 280 compounds and tested them against MRSA grown in a laboratory dish, allowing them to identify two, from the same class, that appeared to be very promising antibiotic candidates. In tests in two mouse models, one of cutaneous MRSA infection and one of systemic MRSA infection, each of those compounds reduced the MRSA population by a factor of 10.
The experiments revealed that the compounds appear to kill bacteria by disrupting their ability to maintain an electrochemical gradient across their cell membranes. This gradient is necessary for many critical cellular functions, including the ability to produce ATP (molecules that cells use to store energy). An antibiotic candidate that Collins' lab discovered in 2020, halicin, appears to work by a similar mechanism, but is specific to gram-negative bacteria (bacteria with thin cell walls). MRSA is a Gram-positive bacteria, with thicker cell walls.
“We have pretty strong evidence that this new structural class is active against Gram-positive pathogens by selectively dissipating the proton motive force in bacteria,” Wong says. “The molecules selectively attack bacterial cell membranes, in a way that does not cause substantial damage to human cell membranes. “Our substantially augmented deep learning approach allowed us to predict this new structural class of antibiotics and led to the discovery that it is not toxic to human cells.”
The researchers have shared their findings with Phare Biography, a nonprofit organization started by Collins and others as part of the Antibiotics-ai Project. The nonprofit now plans to conduct a more detailed analysis of the chemical properties and potential clinical use of these compounds. Meanwhile, Collins' lab is working on designing additional drug candidates based on the new study's findings, as well as using the models to search for compounds that can kill other types of bacteria.
“We are already taking advantage of similar approaches based on chemical substructures to design compounds de novo, and of course, we can easily adopt this off-the-shelf approach to discover new classes of antibiotics against different pathogens,” Wong says.
In addition to MIT, Harvard, and the Broad Institute, institutions contributing to the paper include Integrated Biosciences, Inc., the Wyss Institute for Biologically Inspired Engineering, and the Leibniz Institute for Polymer Research in Dresden, Germany. The research was funded by the James S. McDonnell Foundation, the US National Institute of Allergy and Infectious Diseases, the Swiss National Science Foundation, the Banting Scholarship Program, the Volkswagen Foundation, the Threat Reduction Agency of Defense, the US National Institutes of Health and the Broad Institute. The Antibiotics-ai project is funded by Audacious Project, Flu Lab, Sea Grape Foundation, Wyss Foundation and an anonymous donor.
