Using machine learning technology, scientists at MIT have found a new antibiotic compound that can kill drug-resistant super bacteria.

Superbugs be buggin’

You’ve probably heard or read about new strains of bacteria aka superbugs that mutate and transform to develop resistance to antibiotics. If we don’t develop new antibiotics in time, it is estimated that 10M people worldwide could be killed annually by these superbugs by 2050. Not a rosy scenario.

To make matters worse, big pharmas have been slow to put their R&D money on battling the superbugs because, of course, antibiotics don’t bring in a lot of $ when compared to other drugs. Why would they spend money on antibiotics, which people usually take only for a few days or a week max, when they can peddle addictive opiates that would make people come back to them for life?

The de-bugging

The MIT team developed a machine learning algorithm and “trained” the model on 2,500 molecules to make it look for chemical features that would be effective at killing bacteria like E.Coli. Once the algorithm was out of training mode, the team let it dissect about 6,000 different compounds from a chemical library.

Through this process, the team was able to identify one molecule which had a chemical structure different from any existing antibiotics that showed strong antibacterial activity while at the same time have low toxicity to human cells. The molecule, named Halicin, is effective at killing deadly drug-resistant strains of bacteria such as Clostridium difficileA. baumannii, and Mycobacterium tuberculosis.

The researchers applied an ointment containing Halicin on mice infected with A. baumannii, a super bacterium that has infected thousands of U.S. soldiers stationed in Iraq and Afghanistan, and the Halicin ointment completely cleared the infections within 24 hrs. In another test, E.Coli did not develop any resistance to Halicin over a 30-day treatment – an amazing feat considering the bacteria becomes 200x more resistant to existing antibiotics over the same period of time.

The power of machine learning

In addition to Halicin, the MIT team is working hard to discover more molecules that’d be effective at fighting superbugs. They’ve let the algorithm go through more than 100M molecules and identified 8 molecule candidates that showed promising antibacterial activity. The time the model took to screen over 100M molecules? Only 3 days.

The algorithm hard at work

This means, machine learning tech is not only effective at finding new antibiotics, but it’ll also cut down the development costs for the new drugs significantly – so no more R&D cost excuses for the big pharmas.

As you guys already know, we’re not really fans of A.I. tech (hard to be when even Elon Musk thinks it could be more dangerous than nukes) but when used right, it def has the potential to save more lives. And not many people know the positive potential of A.I. better than Regina Barzilay, one of the leaders of the MIT team. Before she worked on the Halicin study, she developed an A.I. technology that can predict breast cancer up to 5 years in advance after diagnosed with breast cancer herself. You’re a real one Dr. Barzilay.

Share this