The chances of finding an antimicrobial chemical against MRSA were 1 in 6 million.
It’s a practically invisible (and notoriously unbeatable) enemy. MRSA, or methicillin-resistant Staphylococcus aureus poses a significant public health risk due to it’s resistance to common antibiotics.
As a result, vulnerable patients who pick up the bug can end up with severe infections with a high risk of mortality, particularly in healthcare settings. It’s estimated that MRSA is directly responsible for over 10,000 deaths a year in the U.S. alone.
A recent article published in the prestigious scientific journal, Nature, has scientists using an unconventional tool to fight MRSA: artificial intelligence.
There are a seemingly endless array of potential chemical structures in the microbiologists’ toolbox, but only a small subset of these have the right properties to combat MRSA.
In the researchers’ study, they started by testing about 40,000 chemicals for their antibiotic activity against MRSA. This formed the foundation — providing critical insights on which compounds showed antibacterial activity against MRSA.
Then, they leveraged this data to train deep learning models, essentially teaching artificial intelligence to recognize chemical structures associated with MRSA antimicrobial activity.
Once trained, the models were then applied to a collossal database of 12 million commercially available chemical compounds. Testing these the ‘old-fashioned’ way would be far too costly and taken way too long to extract meaningful insights.
But the deep learning model sifted through these 12 million compounds with ease, allowing the researchers to efficiently identify potential antibiotic candidates from a vast pool of possibilities.
Of the 12 million, only 2 compounds were found to be effective in killing MRSA in lab dishes and mouse models of infections. These compounds worked by disrupting the outer layer of the bacteria known as the cell membrane, impairing critical functions like energy production.
Promisingly, the new drug candidates were found to have low toxicity against human cells.