Antibiotics are the most effective weapon against bacterial infections, but mainly as a result of excessive use, resistant bacteria have become an increasing problem. Bacterial resistance to antibiotics is classified by the WHO as one of the biggest threats to human health and is sometimes called a “silent pandemic”.
Now researchers at, among others, MIT in the US have used AI to screen out new types of compounds that attack resistant bacteria.
AI washed
In an article in the journal Nature they describe how they let AI come up with suggestions for chemical structures that they then tested in the lab and in animal experiments. In addition, the researchers managed to understand how the AI arrived at which molecules could become new effective antibiotics.
In the study, the researchers focused on methicillin-resistant Staphylococcus aureus (MRSA). MRSA can cause everything from skin infections to life-threatening blood infections.
The AI models first analyzed around 12 million different molecules and, using deep learning, they picked out candidates that appeared to be both effective against bacteria and harmless to human cells.
Affects cell membranes
The researchers then refined the search and used additional deep learning models. In total, they found 280 different molecules that they went ahead with and tested in the lab. They saw that two of these gave promising results against MRSA in particular. In a next step, they were tested on animals infected with MRSA and they proved to be effective.
Through experiments, they were able to see that the new antibiotic models seem to knock out the bacteria by affecting the cell membranes.
The researchers now want to go further and investigate whether it is possible to make clinically useful drugs based on the new discoveries.