MIT researchers use machine learning to pinpoint molecule capable of killing drug-resistant bacteria in approach said to usher in a ‘new age’
Scientists have used artificial intelligence to discover a powerful new antibiotic in what they called a world first.
MIT researchers used a new machine learning model they said was able to screen more than 100 million molecules in only three days.
The AI was used to target the problem of drug resistance, in which bacteria can no longer be treated by existing antibiotics.
The use of AI in drug discovery is already well-established, but researchers at MIT said their new machine-learning model was significantly more accurate.
‘In silico’ tests
Such models are used to carry out “in silico” drug screening, which promises to significantly reduce the cost of finding new medicines.
Testing costs have until now posed a barrier to finding new antibiotics that can tackle resistant bacteria.
In this case, researchers trained their machine learning model on about 2,500 molecules, giving it the ability to pinpoint certain traits, such as low toxicity and the ability to kill E. coli.
The model then scanned a library of about 6,000 compounds, and picked one out that it predicted to have strong antibacterial activity and which had a chemical structure different from any existing antibiotics.
The molecule had previously only been investigated for use with diabetes sufferers, but researchers found it was able to kill many bacteria resistant to other known treatments.
In tests on mice, for instance, it cleared infections of A. baumannii, a bacterium resistant to all known antibiotics.
MIT’s team also used the AI to screen more than 100 million molecules from another database of about 1.5 billion compounds, a process that took only three days and identified 23 other candidates for new antibiotics.
They are carrying out further tests on eight of the molecules that appeared particularly promising.
“We wanted to develop a platform that would allow us to harness the power of artificial intelligence to usher in a new age of antibiotic drug discovery,” said researcher James Collins of MIT’s Institute for Medical Engineering and Science (IMES) and Department of Biological Engineering, and one of the senior authors of the study.
“Our approach revealed this amazing molecule which is arguably one of the more powerful antibiotics that has been discovered.”
Regina Barzilay, of MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), the study’s other senior author, said the model was capable of exploring “large chemical spaces” that can be “prohibitively expensive” for conventional experimental approaches.