Researchers have used Artificial intelligence to kill Methicillin-Resistant Staphylococcus aureus (MRSA)
They have utilised artificial intelligence to identify a class of compounds capable of combating MRSA, a drug-resistant bacterium responsible for over 10,000 deaths annually in the United States.
Antibiotics AI project
The research, part of the Antibiotics-AI Project at MIT, showcases the potential of artificial intelligence (AI) in drug discovery and provides valuable insights into the mechanisms underlying antibiotic potency.
MRSA, infecting more than 80,000 individuals in the U.S. each year, poses a significant health threat, causing skin infections, pneumonia, and, in severe cases, life-threatening sepsis.
Traditional approaches to finding antibiotics face challenges in addressing drug-resistant strains, prompting researchers to turn to innovative technologies like deep learning.
The study’s lead authors, Felix Wong and Erica Zheng, used deep learning models trained on an expanded dataset of approximately 39,000 compounds to predict antibiotic activity against MRSA.
The researchers then utilised a Monte Carlo tree search algorithm to decipher the factors influencing the model’s predictions. This breakthrough allowed them to gain insights into the chemical structures associated with antimicrobial activity, transforming the previously opaque “black box” nature of deep learning models.
Designing more effective antibiotics
One of the key innovations of the study is the ability to explain the information used by the AI model in making antibiotic potency predictions. James Collins, the Termeer Professor of Medical Engineering and Science at MIT, emphasised the importance of understanding the underlying features guiding the model’s predictions.
This knowledge acts as a foundation for designing more effective antibiotics, surpassing the capabilities of the compounds initially identified by the AI.
The researchers screened a vast library of around 12 million commercially available compounds, identifying five different classes with promising antimicrobial activity against MRSA.
Utalising AI
To further refine their selection, three additional deep learning models predicted the compounds’ toxicity to human cells. The integration of these models led to the identification of compounds demonstrating potent antimicrobial effects with minimal toxicity against human cells.
Out of the approximately 280 compounds tested, two from the same class emerged as highly promising antibiotic candidates. In mouse models of MRSA infection, these compounds significantly reduced the MRSA population, showcasing their potential efficacy in vivo.
The compounds seem to disrupt bacterial cell membranes, selectively targeting MRSA without causing substantial damage to human cell membranes.
What does this mean for the future?
The researchers have shared their findings with Phare Bio, a nonprofit associated with the Antibiotics-AI Project, which plans to conduct detailed analyses of the compounds for potential clinical use.
MIT researchers are actively designing additional drug candidates based on the study’s findings and leveraging similar approaches to target different bacterial pathogens.
The institutions involved in the collaborative effort include MIT, Harvard, the Broad Institute, Integrated Biosciences Inc, the Wyss Institute for Biologically Inspired Engineering, and the Leibniz Institute of Polymer Research in Dresden, Germany.
The study not only marks a significant advancement in the fight against drug-resistant bacteria but also underscores the potential of AI-driven drug discovery in revolutionising the field of medicine.
