Medicosnext
MIT researchers have employed deep learning to identify a class of compounds effective against drug-resistant bacteria, particularly methicillin-resistant Staphylococcus aureus (MRSA), responsible for over 10,000 deaths annually in the United States. The study published in Nature demonstrated the compounds’ potency in lab dishes and mouse models, emphasizing their low toxicity to human cells. The deep-learning model revealed the information it used for antibiotic potency predictions, aiding researchers in designing more effective drugs. By screening about 12 million compounds, the researchers identified five classes predicted to be active against MRSA, with two promising candidates reducing MRSA populations in mouse models. Initially, the scientists trained an advanced deep learning model using significantly enlarged datasets. This training data was created through the examination of approximately 39,000 compounds to assess their antibiotic activity against MRSA. The compounds selectively attacked bacterial cell membranes without causing substantial damage to human cell membranes. The study contributes to the Antibiotics-AI Project at MIT, aiming to discover new antibiotics against deadly bacteria over seven years. The researchers will further analyze the chemical properties and clinical potential of these compounds, while continuing efforts to design additional drug candidates targeting various bacteria types. The Antibiotics-AI Project is funded by the Audacious Project, Flu Lab, the Sea Grape Foundation, the Wyss Foundation, and an anonymous donor.
Reference: https://news.mit.edu/2023/using-ai-mit-researchers-identify-antibiotic-candidates-1220
