With nearly 125 billion potential side effects from different drug combinations, a new artificial intelligence system developed at Stanford wants to better predict these complications.
Called Decagon, the AI tool can help predict likely effects from drug pairings, helping doctors make better decisions in drug prescription and aiding researchers in creating better drug combinations to treat illness.
Identifying Patterns In Drug Interaction Side Effects
To accurately determine the impact of various drug combinations remains a difficult task. Today, there are 1,000 different known side effects along with 5,000 drugs on the market, and a majority of them have never been prescribed together.
“It’s practically impossible to test a new drug in combination with all other drugs,” computer science postdoc fellow Marinka Zitnik said in a statement, “because just for one drug that would be 5,000 new experiments.”
Zitnik worked with Jure Leskovec, an associate professor of computer science, and colleagues to come up with Decagon, which they presented July 10 at the 2018 meeting of the International Society for Computational Biology in Chicago.
First, the team studied how drugs affect the body’s underlying cellular machinery composed by a huge network, detailing how over 19,000 proteins in the body interact with one another and how various drugs affect such proteins.
They then used over 4 million of the 125 billion known associations between drugs and side effects to come up with a deep learning system that identifies patterns in the emergence of side effects from how drugs target proteins.
Deep learning, a type of AI modeled after the brain, pores over complex data and extracts patterns in the data that are often abstract and even counterintuitive. In this case, the system infers patterns about drug interaction side effects and then predicts formerly unidentified consequences from taking multiple drugs together.
AI And Machine Learning At Work
For instance, despite no earlier indication that they would create muscle inflammation, the cholesterol medication atorvastatin paired with hypertension drug amlipidine was predicted by the AI tool to do so. The team later confirmed that their prediction was previously suggested in a 2017 case study.
The team plans to expand the work to comprise more complex regimens as well as produce a more user-friendly tool for physicians. If drug side effects today are mostly discovered by accident, their approach can lead to greater safety, Leskovec said.
Dubbed polypharmacy, the practice of combining drugs proves to be a real threat. In the last month alone, 23 percent of Americans took at least two prescription drugs based on an estimate from the CDC.
Experts have touted AI-powered analytics and tools in preventing life-threatening patient complications. McKinsey analysts, for instance, estimated that machine learning algorithms could help medicine and pharmaceutical firms save up to $100 billion annually.
