Computer models successfully predict drug side effects

New computer models were able to successfully predict negative side effects for hundreds of currently marketed drugs, report researchers from the UCSF School of Pharmacy, SeaChange Pharmaceuticals, and Novartis Institutes for Biomedical Research in a paper published online this week in the journal Nature.

The effort could set the stage for the mass screening of potential new drug compounds for difficult-to-predict yet surprisingly common off-target effects, in which a drug meant to bind to one protein to treat a disease instead binds to others, producing adverse and even dangerous side effects. Such unforeseen adverse effects are the second most common reason, after lack of efficacy, that drugs fail in clinical trials.

The scientists ran a computer screen on 656 drugs that are currently in clinical use to predict their activity on 73 “unintended side-effect targets” from a Novartis safety panel used to test drugs for adverse reactions. About half of the computer-predicted off-target bindings were confirmed via proprietary databases or new experiments.

Hundreds of compounds have never been tackled at once before, according to Brian Shoichet, PhD, a faculty member in the School of Pharmacy’s Department of Pharmaceutical Chemistry, who was joint advisor on the project with Laszlo Urban, MD, PhD, of Novartis.

“Someday, when you’re deciding among hundreds of thousands of compounds to pursue [in drug development], you could run a computer program to prioritize for those that may be safest,” said Michael Keiser, PhD, co-first author of the paper, who started working on the project as a doctoral student in Shoichet’s lab. Upon his graduation, co-founded SeaChange with Shoichet and John Irwin, PhD, also of UCSF.