The system analyses how proteins in the human body are affected by the roughly 5,000 pharmaceuticals on the market today. It uses more than four million known associations between drugs and side effects, then applies machine learning to identify patterns and predict how two drugs will interact. According to the Stanford team, Decagon could be an invaluable tool for doctors, as side effects from multiple drugs are often only discovered once the patient has reported them.
"It's practically impossible to test a new drug in combination with all other drugs, because just for one drug that would be five thousand new experiments," said Marinka Zitnik, a postdoctoral fellow in computer science at Stanford. With some new drug combinations, she said, "truly we don't know what will happen."
To test the accuracy of the system, the team looked at predictions made by Decagon to see if they matched up with actual medical discoveries. For example, there was no indication in the team's original data that the combination of atorvastatin, a cholesterol drug, and amlodipine, a blood pressure medication, would lead to muscle inflammation. However, the AI system predicted that it would, and it turned out to be correct. Although it did not appear in the original data, a case report from 2017 suggested the drug combination had led to a dangerous kind of muscle inflammation.
When the researchers explored the medical literature for evidence of 10 side effects predicted by Decagon, they found that five out of the 10 have recently been confirmed, lending further credence to the system’s predictions.
"It was surprising that protein interaction networks reveal so much about drug side effects," said Jure Leskovec, an associate professor of computer science at Stanford. "Today, drug side effects are discovered essentially by accident, and our approach has the potential to lead to more effective and safer healthcare."
For now, Decagon only considers side effects associated with pairs of drugs, but in the future the team hopes to extend the analysis to include more complex combinations.