Researchers in Canada and the United States have used deep learning to derive an antibiotic that can attack Acinetobacter baumannii, a resistant bacterium that infects wounds and causes pneumonia. A paper published in Nature Chemical Biology describes how researchers used training data to measure the effects of known drugs against tough bacteria, according to the BBC. The learning algorithm then predicted the efficacy of 6,680 compounds for which there was no data on efficacy against bacteria.
The program lasted an hour and a half and narrowed the list down to 240 promising candidates. Lab tests have shown that nine of these are effective, and one, now called Abausin, is extremely potent. Testing 240 compounds in a lab sounds like a lot of work, but it’s more effective than testing nearly 6,700 compounds.
Interestingly, the new antibiotics appear to be effective only against target microbes, which is a plus. It’s not available to people yet, and it may not be available for some time – drug testing is what it is. But it’s still a great example of how machine learning can enhance human intelligence, allowing scientists and others to focus on what really matters.
The WHO has identified Acinetobacter baumannii as one of the major superbugs threatening the world, and development of a weapon against it would be very welcome. This technology is expected to significantly reduce the time it takes to develop new drugs. I also sometimes wonder if there are other areas where AI techniques can quickly pick out alternatives and free humans to focus on more promising candidates.
Want to learn about machine learning algorithms? Google can help. Alternatively, you can dive into longer courses.
