Targeting MRSA with Deep Learning
MIT researchers, employing deep learning, have uncovered a novel class of compounds with the potential to combat the drug-resistant bacterium Methicillin-resistant Staphylococcus aureus (MRSA). MRSA is responsible for more than 10,000 deaths annually in the United States due to deadly infections.
Groundbreaking Antibiotic Potency Predictions
Published in Nature, the study highlights the compounds’ efficacy against MRSA in lab dishes and mouse models, showcasing low toxicity against human cells. The groundbreaking aspect is the transparency in the deep-learning model’s predictions, providing insights into the features influencing antibiotic potency.
Decoding the Black Box: Unraveling AI Predictions
The research team, led by James Collins, Termeer Professor of Medical Engineering and Science at MIT, addressed the “black box” nature of deep-learning models. By training the model on extensive datasets and implementing the Monte Carlo tree search algorithm, they deciphered the molecular features influencing the predictions.
Screening Millions of Compounds: A Multifaceted Approach
The researchers employed three additional deep-learning models to assess compounds’ toxicity against human cells. By combining this data with antimicrobial predictions, they screened 12 million commercially available compounds. Five distinct classes, based on chemical substructures, emerged as potential candidates against MRSA.
From Prediction to Validation: Identifying Promising Antibiotics
Out of the 280 compounds tested, two from the same class demonstrated promising antibiotic potential, reducing MRSA populations significantly in mouse models. The compounds disrupt bacterial cell membranes, selectively targeting Gram-positive pathogens without substantial damage to human cells.
Future Steps: Analysis and Design
The findings have been shared with Phare Bio, a nonprofit associated with the Antibiotics-AI Project. Further analysis of chemical properties and clinical potential is underway. Collins’ lab aims to design additional drug candidates based on the study’s outcomes and explore compounds effective against different bacterial pathogens.
Author Introduction: Pritish Kumar Halder
Pritish Kumar Halder is a seasoned researcher with a passion for cutting-edge advancements in medical engineering and science. With a keen interest in AI applications in healthcare, Halder has been contributing valuable insights to the field. As an advocate for transparent and innovative research, Halder continues to explore the intersections of technology and medicine.
