Abaucin: an AI-found antibiotic against a hospital bacterium

Why is A. baumannii such a hard bacterium to treat?

Acinetobacter baumannii is a bacterium found mostly in hospitals, where it can infect patients who are already frail. The authors describe it as a nosocomial pathogen (meaning linked to healthcare) that often shows resistance to several antibiotics at once.

Finding new antibiotics against it has proved difficult with conventional screening methods, which test a huge number of molecules one by one.

The authors note that bacteria like this one are a real challenge because, with few options left, each new active molecule is precious.

How did AI help find a new molecule?

The team first tested about 7,500 molecules in the laboratory to see which ones slowed the bacterium's growth. These results were used to train a neural network, an AI program that learns to link a molecule's shape to its effect.

The model then scanned, by computer, molecules with new structures, to estimate which might act against the bacterium. This fast exploration of "chemical space" raises the chance of finding new antibacterial molecules. This is how abaucin was found.

The key advantage is speed. Instead of testing every molecule in the laboratory, the AI can narrow the list to the few that are most worth testing.

What did the experiments on abaucin show?

Abaucin is a narrow-spectrum antibacterial compound: its activity targets A. baumannii. The researchers showed that it disturbs the trafficking of lipoproteins, molecules that the bacterium moves around to build its outer layer, through a mechanism involving a protein called LolE.

In a mouse wound model, abaucin was also able to control an A. baumannii infection.

In other words, the team did not only find a molecule that works: it also began to understand how it works, which is useful for improving it.

What could this mean for patients?

For now, nothing for patients: abaucin is not a medicine. It is presented as a promising lead, meaning a starting point to be improved.

The value of this work is to show that an AI can help explore the space of molecules faster to find antibiotics against difficult bacteria. A narrow-spectrum antibiotic aims at a single bacterium, which in principle spares the others.

What are the limits of this study?

The results come from the laboratory and from mice. Many steps remain before a possible medicine: improving the molecule, checking its safety, testing it in animals on a larger scale, then in people. All of this remains to be proven.

A wound infection in mice does not reproduce every real-life situation of patients. And the study covers a single bacterium: nothing can be concluded about others.

The first test in an animal is encouraging, but a single mouse model cannot tell us how the molecule would behave in a person, how safe it is, or how bacteria might adapt to it.