In a study published in Nature, MIT researchers used deep learning to identify a new class of antibiotic candidates that can kill methicillin-resistant Staphylococcus aureus (MRSA), a drug-resistant bacterium responsible for more than 10,000 deaths annually in the United States. This approach allowed them to sift through millions of other compounds, generating predictions of which ones may have potent antimicrobial properties.
Identifying Promising Antibiotic Candidates
The researchers trained a deep learning model on the known structures and antimicrobial activities of hundreds of thousands of molecules. They then used this model to screen a library of over 6 million compounds, identifying 23 candidates predicted to have potent antimicrobial effects against MRSA. Of these, 8 were previously unknown antibiotics.
Validating the Predictions
To validate the model’s predictions, the researchers tested the 23 candidate compounds in the laboratory. They found that 8 of the compounds were able to kill MRSA effectively, including 6 that had not been previously reported as antibiotics. Further testing showed that these new antibiotic candidates were effective against a range of other drug-resistant bacteria as well.
The ability of this deep learning approach to rapidly identify new antibiotic candidates from millions of compounds holds promise for accelerating the discovery of much-needed new antimicrobial therapies to combat the growing threat of drug-resistant infections.