AI is being used to understand the interactions of multiple drugs

Millions of people take 5 or more medications a day, but testing the many side-effects of those pharmaceutical drug combinations has historically been difficult. Its difficult to trace back a cause of why a certain effect is happening when many variables are introduced. Researchers at Stanford Univ. have invented a way to predict side-effects using computer modeling based-on artificial intelligence. The team explained that most drug combinations (called poly-pharmacy) have never been systematically studied.

Their AI software system is called Decagon, and they say it can help physicians make better decisions about what drugs to prescribe. It could also help researchers find better combinations of prescription drugs to treat complex diseases. This is a practical example of how medicine is being revolutionized by breakthroughs in machine learning and AI. With so many prescription drugs currently on the pharmaceutical market: “it’s practically impossible to test a new drug in combination with all other drugs — because just for one drug that would be 5,000 new experiments,”

“With some new drug combinations we don’t know what will happen.” The researchers explained that poly-pharmacy side-effects happen because of drug -to- drug interactions, the effects of one pharmaceutical drug may change (positively or negatively) if it’s taken with another drug, leading to a lot of unpredictable complexity.

The knowledge of medical drug interactions is often limited, because the complex relationships between many pharmaceutical combinations is only rarely observed. Discovering poly-pharmacy side-effects is a serious challenge that’s important for patient health.

The most important thing to understanding this is tracking how pharmaceutical drugs affect proteins. The researchers created a data-base containing descriptions of how 19,000+ proteins found in the human body interact with each other, and how various drugs affect these proteins. Using 4+ million known associations between drugs and their side-effects, the team crafted a method to identify patterns in how side-effects arise — based-on the way pharmaceutical drugs interact with proteins.

With that method, the system could predict the outcome of taking 2 drugs together. To evaluate their method, the group looked to see if its predictions came true. In many cases, they did. For example: there was no existing indication that the combination of a cholesterol drug named Lipitor and a blood-pressure medication called Norvasc, could lead to muscle inflammation. But Decagon predicted that it would, and that foresight was proven correct.

The researchers are already looking ahead to build on this technique. The team hopes to extend their results to include more multiple drug interactions. They aim to create a user-friendly tool, that gives physicians guidance on whether it’s a good idea to prescribe a particular pharmaceutical drug to a particular patient. And to help researchers developing drug protocols for complex diseases, with fewer side-effects.