Neural networks trace patterns and finds efficient antibiotics.
A new antibiotic drug has been named after HAL 9000 from the film 2001: A Space Odyssey, in reference to the AI network that discovered it. Halicin as it is now dubbed, was originally designed for diabetics patients, yet failed to pass clinical trials.
A recent paper published by MIT researchers, describes a breakthrough methodology for finding effective antibiotics with the assistance of a deep learning algorithm. They initially curated a list of approximately 2300 molecules known for their ability to harm bacteria and put them into the network. The researchers purposely left the antibiotics unclassified and let the neural network conclude what the molecules had in common and discover significant patterns autonomously.
Once the network has been trained on the provided data, the researchers fed it a new list of 6000 molecules from medicines that had been developed for different purposes but were found to be ineffective (including our hero halicin). The algorithm then attempted to “guess” which medicines would be effective antibiotics, particularly attempting to identify the ones with molecules different toa known medicine.
The first hundred matches were physically tested in the laboratory for effectiveness, and among them were completely new antibiotics. Among those, halicin has been found to be effective in a wide spectrum to attack various bacteria. As long as halicin is nothing like its namesake when it comes to human life – and only sabotages the bacteria it is designed to attack, we should be safe 🙂
With bacteria strains developing significant antibiotic resistance, such a discovery is rather important. Current predictions indicate that resistant infections could cause up to ten million fatalities per year by 2050.
The hope is that from this breakthrough scientists can establish a comprehensive list of the hundreds of millions of candidates for effective medication. Machine Learning algorithms will then be able to identify the successful candidates antibiotics, which target and attack a specific bacterium while leaving the good bacteria surrounding it intact.
And what about COVID-19, AKA Coronavirus? Well, virus vaccination is a whole different game. Maybe we will hear about a data-driven approach that helps bring a cure to that, sooner than later.