Researchers at MIT have identified a family of drugs that can eradicate a drug-resistant bacteria that kills over 10,000 people annually in the US by using a kind of artificial intelligence called deep learning. Researchers demonstrated that these substances could eradicate methicillin-resistant Staphylococcus aureus (MRSA) cultivated in a lab dish and two mice models of MRSA infection in a report published in Nature. The chemicals are especially promising for drugs since they also exhibit very little toxicity against human cells.
The fact that the researchers were able to determine the kind of data the deep-learning model utilized to forecast the antibiotic potency is a significant breakthrough of the current work. With this information, scientists may be able to create new medications that function even better than the ones the model suggested.
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 is time-efficient, resource-efficient, and mechanistically insightful, from a chemical-structure standpoint, in ways that we haven’t had to date.
James Collins, the Termeer Professor of Medical Engineering and Science
The project is part of the Antibiotics-AI Project at MIT, and its primary authors are 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 mentored by Collins. Under Collins’ direction, the goal of this seven-year study is to identify novel classes of antibiotics that can combat seven different kinds of harmful bacteria.
Predictions (Explainable)
More than 80,000 Americans get MRSA each year, and the bacteria frequently results in pneumonia or skin infections. Sepsis, a potentially lethal bloodstream infection, can result from severe instances.
Collins and his associates at the Abdul Latif Jameel Clinic for Machine Learning in Health (Jameel Clinic) at MIT have been utilizing deep learning to look for novel medicines for the last several years. Their research has produced several drug-resistant bacteria, including Acinetobacter baumannii, which is frequently seen in hospitals, as well as possible medications.
Deep learning methods that can recognize chemical structures linked to antibacterial activity were used to identify these substances. After sorting through millions of more chemicals, these algorithms forecast which ones could have potent antibacterial action.
What we set out to do in this study was to open the black box? These models consist of very large numbers of calculations that mimic neural connections, and no one really knows what’s going on underneath the hood.
Felix Wong
Using far larger datasets, the researchers first constructed a deep-learning model. 39,000 compounds were tested for antibiotic efficacy against MRSA to produce the training data, which was then input into the model along with details on the chemical structures of the compounds.
You can represent basically any molecule as a chemical structure, and also you tell the model if that chemical structure is antibacterial or not. The model is trained on many examples like this. If you then give it any new molecule, a new arrangement of atoms and bonds, it can tell you a probability that that compound is predicted to be antibacterial.
Felix Wong
The researchers utilized a method called Monte Carlo tree search, which has been used to help make other deep learning models, like AlphaGo, more explainable, to find out how the model was generating its predictions. According to this search technique, the model can forecast which molecule’s substructures are most likely responsible for the activity, in addition to providing an estimate of each molecule’s antimicrobial activity.
Potent activity
The researchers trained three more deep-learning models to predict whether the chemicals were hazardous to three distinct types of human cells to further reduce the number of potential medications. Through the integration of this data with the anticipated antibacterial activity, the scientists identified substances capable of eliminating microorganisms with negligible negative consequences on human health.
To further narrow down the pool of possible drugs, the researchers trained three more deep-learning models to predict if the compounds were harmful to three different types of human cells. By combining this information with the predicted antibacterial activity, the researchers were able to identify compounds that may eradicate germs while having less detrimental effects on human health.
After evaluating over 280 compounds against MRSA cultured in a lab dish, the researchers were able to find two, that belonged to the same class and seemed to be extremely promising antibiotic candidates. Each of those chemicals lowered the MRSA population by a factor of ten in experiments conducted on two animal models: one for an MRSA cutaneous infection and the other for an MRSA systemic infection.
Also Read| Morning sickness in pregnant women is caused by a protein (GDF15) – Research
According to experiments, the substances seem to kill bacteria by interfering with their capacity to keep an electrochemical gradient across their cell membranes. Numerous essential cell processes, such as the synthesis of ATP (molecules that cells utilize to store energy), depend on this gradient. Halicin, an antibiotic candidate identified by Collins’ group in 2020, appears to function by a comparable mechanism but is limited to Gram-negative bacteria (bacteria with thin cell walls). Gram-positive bacteria with thicker cell walls are called MRSA.
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. The molecules are attacking bacterial cell membranes selectively, in a way that does not incur substantial damage in human cell membranes. Our substantially augmented deep learning approach allowed us to predict this new structural class of antibiotics and enabled the finding that it is not toxic against human cells.
Felix Wong
As part of the Antibiotics-AI Project, Collins and other creators founded the nonprofit organization Phare Bio, with which the researchers have shared their discoveries. The charity now intends to conduct a more thorough examination of these substances’ chemical characteristics and their medical applications. In the meanwhile, Collins’ group is utilizing the models to uncover chemicals that can kill other kinds of bacteria and is working on creating more therapeutic candidates based on the results of the recent study.
Also Read| Harmful immune responses can be prevented by redesigning future mRNA-based therapeutics – Research
We are already leveraging similar approaches based on chemical substructures to design compounds de novo, and of course, we can readily adopt this approach out of the box to discover new classes of antibiotics against different pathogens
Felix Wong
Source: MIT News
Journal Reference: Wong, F., Zheng, E. J., Valeri, J. A., Donghia, N. M., Anahtar, M. N., Omori, S., Li, A., Krishnan, A., Jin, W., Manson, A. L., Friedrichs, J., Helbig, R., Hajian, B., Fiejtek, D. K., Wagner, F. F., Soutter, H. H., Earl, A. M., Stokes, J. M., Renner, L. D., . . . Collins, J. J. (2023). Discovery of a structural class of antibiotics with explainable deep learning. Nature, 1-9. https://doi.org/10.1038/s41586-023-06887-8
Last Updated
