Can AI predict adverse events from new drug combinations?

By Erik L. Ridley, The Science Advisory Board contributing writer

April 8, 2022 -- A new artificial intelligence (AI) model may be able to help clinicians predict if new combinations of drugs will produce side effects that are worse than their potential benefit, according to research presented at this week’s annual meeting of the American Association for Cancer Research (AACR).

In a proof-of-concept project, a Dutch research team developed a deep-learning algorithm that was able to predict adverse drug events from new therapy combinations. They shared their results in a poster presentation at the New Orleans meeting.

"Our approach can help us understand the relationship between the effects of different drugs in relation to the disease context," said senior author Bart Westerman, PhD, an associate professor at the Cancer Center Amsterdam, in a statement from the AACR.

In their effort to apply AI to better predict adverse drug events from new combination therapies, Westerman, Aslı Küçükosmanoğlu (presenter and graduate student), and colleagues first gathered data from the U.S. Food and Drug Administration Adverse Event Reporting System (FAERS). They then grouped events that frequently occur together in order to simplify the analysis and strengthen the associations between a drug and its side-effect profile.

Using this data, the researchers then trained a convolutional neural network to identify common patterns between drugs and their side effects. After testing their model on unseen adverse event profiles of combination therapies, the researchers found that their algorithm -- called the adverse events atlas -- recognized these new profiles. This demonstrated that the AI could easily predict adverse effects of combination therapy, according to Westerman, et al.

"We were able to determine the sum of individual therapy effects through simple algebraic calculation of the latent space descriptors," Westerman said. "Since this approach reduces noise in the data because the algorithm is trained to recognize global patterns, it can accurately capture the side effects of combination therapies."

In the next phase of their work, the researchers are now developing a statistical approach to quantify the model's accuracy.

"Given that the landscape of drug interactions is highly complex and involves many molecular, macromolecular, cellular, and organ processes, it is unlikely that our approach will lead to black-and-white decisions," Westerman said. "The adverse events atlas is still in the proof-of-concept phase, but the most important finding is that we were able to get snapshots of the interplay of drugs, diseases, and the human body as described by millions of patients."


Copyright © 2022 scienceboard.net
 

Create an Account

Already have an account? Sign in Here

To access all ScienceBoard content create a free account now:



Email Address:  

First Name:

Last Name:

Learn about ScienceBoard

Get the latest life sciences research and industry news, delivered straight to your inbox, for free.

Why subscribe?

ScienceBoard is uniquely focused on the business of research, addressing the biggest problems that the biomedical industry face. You’ll get breaking news, events coverage, and deep dives into the science that drives innovation, delivered to your inbox daily.

 
I have read and agree to the privacy policy and terms of service and wish to opt-in for ScienceBoard.net.

Email Preferences



Letter from the Editor Please send me twice-weekly roundups of all the latest life research and industry news.
SAB Announcements Please send me the latest announcements from The Science Advisory Board and their partners.
Spotlight Receive notifications about new content, services, or educational resources designed to help you sharpen your skills and grow professionally.