Researchers found a new antibiotic that kills Acinetobacter baumannii (pink), a drug-resistant bacteria, using artificial intelligence.

"Eventually, doctors will adopt AI and algorithms as their work partners. This levelling of the medical knowledge landscape will ultimately lead to a new premium: to find and train doctors with the highest level of emotional intelligence." - Eric Topol.

The researchers found the new drug in a library of nearly 7,000 possible compounds. Then, they trained a machine-learning model to determine if a chemical compound would stop "A.baumannii" from growing.

Drug discovery

Many harmful bacteria have become more immune to antibiotics in the past few decades, but only a few new antibiotics have been made. In the first demo, the scientists taught a machine-learning programme to find chemical structures that could stop E. coli from growing. In a screen of more than 100 million compounds, this algorithm found a molecule that the researchers called "halicin", after the imaginary, artificial intelligence system from "2001: A Space Odyssey." They showed that this molecule could kill not only E. coli but several other types of germs that are hard to treat.

A.baumannii - growth of bacteria

To get the data they needed to train their computer model, the researchers first grew A. baumannii in a lab dish and gave it about 7,500 different chemicals to see which ones could stop the microbe from growing. Then, they told the model of how each molecule was built. 

After the model was trained, researchers used it to look at 6,680 compounds from the Drug Repurposing Hub at the Broad Institute that it had never seen before. This study took less than two hours and turned up a few hundred top hits. Out of these, the researchers picked 240 to test in the lab. Again, they focused on compounds with different structures than existing antibiotics or molecules from the training data.

Abaucin - new medicine

The medicine, which the scientists have termed "abaucin", was effective against "A.baumannii" wound infections in mouse trials. 

Experiments showed that the medication kills cells by disrupting their ability to move proteins from the cytoplasm to the cell membrane, a process called lipoprotein trafficking. In addition, the medicine appears to target and inhibit LolE, a protein critical to this procedure.

Conclusion

Nosocomial Gram-negative pathogen Acinetobacter baumannii frequently exhibits multidrug resistance. Traditional screening methods have not successfully identified novel antibiotics effective against "A.baumannii". The researchers acknowledge the usefulness of machine learning approaches in expediting the rapid exploration of chemical space necessary to identify promising novel antibacterial chemicals. Using this growth inhibition dataset, the researchers trained a neural network to make in silico predictions for structurally novel compounds active against "A.baumannii". 

Furthermore, researchers seek to apply their modelling method to develop medications for drug-resistant Staphylococcus aureus and Pseudomonas aeruginosa infections.

Image source: Unsplash