With the help of Artificial Intelligence (AI) scientists have discovered a new antibiotic. The antibiotic is part of a new drug class and will be used to treat the drug-resistant Staphylococcus aureus, which is more commonly known as MRSA.
MRSA is a superbug, and is notoriously difficult to treat. Finding a treatment that works is vital- currently over 35,000 people a year die from the infection. Whilst many people harmlessly carry the bacteria on their skin, it becomes dangerous when patients in a hospital contract it through open wounds.
Deep learning AI Models
Using a deep-learning algorithm, the AI was able to identify new compounds and to predict how the compounds would work against MRSA. The scientists involved in the discovery were able to observe how the AI systems learned, and apply it to their findings. James Collins, professor of Medical Engineering and Science at the Massachusetts Institute of Technology (MIT) and one of the study’s authors says:
“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.”
Black Box AI
Typically most AI Deep learning systems are referred to as ‘Black Box’, as its reasoning and logic is not available to operators of the AI systems. This is partly due to the complex way that AI systems learn, by processing extremely large datasets and following patterns. This meant that whilst the AI system could tell the researchers what bacteria could be effective the researchers couldn’t understand how the AI system came to this conclusion. In order to get around this issue the researchers adapted an algorithm known as Monte Carlo tree search, which has been used to help make other deep learning models, such as the board game playing AI system AlphaGo, more explainable. This algorithm allows the AI model to generate not only an estimate of each molecule’s antimicrobial activity, but also a prediction for which substructures of the molecule likely account for that activity.
The researchers have passed on their findings to the nonprofit Phare Bio, founded as part of the Antibiotics-AI Project, which will now conduct further analysis of the compounds’ chemical properties and potential clinical applications. Meanwhile, Collins’ lab will continue to design additional drug candidates using the findings of the study, as well as employing the models to search for compounds capable of killing other types of bacteria.
