Using a sort of artificial intelligence often known as deep learning, MIT researchers have discovered a category of compounds that may kill a drug-resistant bacterium that causes greater than 10,000 deaths in america every 12 months.

In a study appearing today in , the researchers showed that these compounds could kill methicillin-resistant (MRSA) grown in a lab dish and in two mouse models of MRSA infection. The compounds also show very low toxicity against human cells, making them particularly good drug candidates.

A key innovation of the brand new study is that the researchers were also capable of work out what kinds of knowledge the deep-learning model was using to make its antibiotic potency predictions. This data could help researchers to design additional drugs which may work even higher than those identified by the model.

“The insight here was that we could see what was being learned by the models to make their predictions that certain molecules would make for good antibiotics. Our work provides a framework that’s time-efficient, resource-efficient, and mechanistically insightful, from a chemical-structure standpoint, in ways in which we haven’t had thus far,” says James Collins, the Termeer Professor of Medical Engineering and Science in MIT’s Institute for Medical Engineering and Science (IMES) and Department of Biological Engineering.

Felix Wong, a postdoc at IMES and the Broad Institute of MIT and Harvard, and Erica Zheng, a former Harvard Medical School graduate student who was advised by Collins, are the lead authors of the study, which is a component of the Antibiotics-AI Project at MIT. The mission of this project, led by Collins, is to find latest classes of antibiotics against seven kinds of deadly bacteria, over seven years.

Explainable predictions

MRSA, which infects greater than 80,000 people in america every 12 months, often causes skin infections or pneumonia. Severe cases can result in sepsis, a potentially fatal bloodstream infection.

Over the past several years, Collins and his colleagues in MIT’s Abdul Latif Jameel Clinic for Machine Learning in Health (Jameel Clinic) have begun using deep learning to try to seek out latest antibiotics. Their work has yielded potential drugs against , a bacterium that is commonly present in hospitals, and lots of other drug-resistant bacteria.

These compounds were identified using deep learning models that may learn to discover chemical structures which can be related to antimicrobial activity. These models then sift through tens of millions of other compounds, generating predictions of which of them can have strong antimicrobial activity.

These kinds of searches have proven fruitful, but one limitation to this approach is that the models are “black boxes,” meaning that there is no such thing as a way of knowing what features the model based its predictions on. If scientists knew how the models were making their predictions, it may very well be easier for them to discover or design additional antibiotics.

“What we got down to do on this study was to open the black box,” Wong says. “These models consist of very large numbers of calculations that mimic neural connections, and nobody really knows what is going on on underneath the hood.”

First, the researchers trained a deep learning model using substantially expanded datasets. They generated this training data by testing about 39,000 compounds for antibiotic activity against MRSA, after which fed this data, plus information on the chemical structures of the compounds, into the model.

“You may represent principally any molecule as a chemical structure, and in addition you tell the model if that chemical structure is antibacterial or not,” Wong says. “The model is trained on many examples like this. In case you then give it any latest molecule, a latest arrangement of atoms and bonds, it may possibly inform you a probability that that compound is predicted to be antibacterial.”

To work out how the model was making its predictions, the researchers adapted an algorithm often known as Monte Carlo tree search, which has been used to assist make other deep learning models, similar to AlphaGo, more explainable. This search algorithm allows the model to generate not only an estimate of every molecule’s antimicrobial activity, but in addition a prediction for which substructures of the molecule likely account for that activity.

Potent activity

To further narrow down the pool of candidate drugs, the researchers trained three additional deep learning models to predict whether the compounds were toxic to a few several types of human cells. By combining this information with the predictions of antimicrobial activity, the researchers discovered compounds that might kill microbes while having minimal hostile effects on the human body.

Using this collection of models, the researchers screened about 12 million compounds, all of that are commercially available. From this collection, the models identified compounds from five different classes, based on chemical substructures inside the molecules, that were predicted to be energetic against MRSA.

The researchers purchased about 280 compounds and tested them against MRSA grown in a lab dish, allowing them to discover two, from the identical class, that gave the impression to be very promising antibiotic candidates. In tests in two mouse models, one in all MRSA skin infection and one in all MRSA systemic infection, each of those compounds reduced the MRSA population by an element of 10.

Experiments revealed that the compounds appear to kill bacteria by disrupting their ability to take care of an electrochemical gradient across their cell membranes. This gradient is required for a lot of critical cell functions, including the flexibility to provide ATP (molecules that cells use to store energy). An antibiotic candidate that Collins’ lab discovered in 2020, halicin, appears to work by an analogous mechanism but is restricted to Gram-negative bacteria (bacteria with thin cell partitions). MRSA is a Gram-positive bacterium, with thicker cell partitions.

“We’ve got pretty strong evidence that this latest structural class is energetic against Gram-positive pathogens by selectively dissipating the proton driving force in bacteria,” Wong says. “The molecules are attacking bacterial cell membranes selectively, in a way that doesn’t incur substantial damage in human cell membranes. Our substantially augmented deep learning approach allowed us to predict this latest structural class of antibiotics and enabled the finding that it shouldn’t be toxic against human cells.”

The researchers have shared their findings with Phare Bio, a nonprofit began by Collins and others as a part of the Antibiotics-AI Project. The nonprofit now plans to do more detailed evaluation of the chemical properties and potential clinical use of those compounds. Meanwhile, Collins’ lab is working on designing additional drug candidates based on the findings of the brand new study, in addition to using the models to hunt compounds that may kill other kinds of bacteria.

“We’re already leveraging similar approaches based on chemical substructures to design compounds de novo, and in fact, we will readily adopt this approach out of the box to find latest classes of antibiotics against different pathogens,” Wong says.

Along with MIT, Harvard, and the Broad Institute, the paper’s contributing institutions are Integrated Biosciences, Inc., the Wyss Institute for Biologically Inspired Engineering, and the Leibniz Institute of Polymer Research in Dresden, Germany. The research was funded by the James S. McDonnell Foundation, the U.S. National Institute of Allergy and Infectious Diseases, the Swiss National Science Foundation, the Banting Fellowships Program, the Volkswagen Foundation, the Defense Threat Reduction Agency, the U.S. National Institutes of Health, and the Broad Institute. The Antibiotics-AI Project is funded by the Audacious Project, Flu Lab, the Sea Grape Foundation, the Wyss Foundation, and an anonymous donor.