“Today, drug side effects are discovered essentially by accident.”

That totally reassuring and absolutely not terrifying fact comes to us courtesy of Jure Leskovec, an associate professor of computer science at Stanford University.  That’s right – when you mix that Panado with Allergex then use a nasal decongestant because it’s freezing outside and you’ve got a headache on top of a blocked nose… yeah, nobody actually really knows how exactly those drugs are interacting in your body and how that might affect you. Isn’t that fun and exciting?! The good news is that there’s an AI in development whose entire job it is to figure out those potential side effects before you spontaneously combust.

Designed by Marinka Zitnik, Monica Agrawal, and Jure Leskovec, Decagon uses a deep learning technique modleled on the functioning of the human brain. The research team chose this technique because it allows Decagon to analyse massive amounts of data and come to conclusions and identify potential outcomes which can be abstract, or seem counter-intuitive.

“[The researchers] composed a massive network describing how the more than 19,000 proteins in our bodies interact with each other and how different drugs affect these proteins,” Science Daily reports. “Using more than four million known associations between drugs and side effects, the team then designed a method to identify patterns in how side effects arise based on how drugs target different proteins.”

On a practical level, Decagon is a necessity in a world where at any given time, many people are taking combinations of over-the-counter drugs, mixing over-the-counter with prescription meds, or even being prescribed combinations by their doctors – all without any real-world way to check how these drugs may affect the body over all. According to the CDC, in June, approximately 23% of Americans took two or more prescription drugs. I imagine the situation wouldn’t be much different in South Africa, and doctors have no real way of knowing how these combinations actually might affect anybody.

Decagon is on track to do exactly what it was designed for, and in many cases the potential side effects it predicted can be found in cases already otherwise reported – leading the team to believe that the rest of its predictions will likely be borne out eventually. Though it currently only predicts potential interactions between pairs of drugs, the research team hope to eventually have Decagon process more complex combinations. They also want to make its interface more user friendly so that physicians will be able to use the tool in future.