Virtual safety panel predicts drugs' side effects
A team of scientists from the UCSF School of Pharmacy, Novartis Institutes for Biomedical Research (NIBR) and SeaChange Pharmaceuticals has developed a set of computer models that can predict negative side effects associated with existing drugs. By speeding up the process and increasing accuracy, the software could potentially save billions in research and decrease the number of animals used in toxicity tests.
The model, based on UCSF’s “similarity ensemble approach” (SEA), uses the similarities between the shape of each drug and thousands of other compounds to predict possible side effects. The theory behind SEA technology is that proteins can be related by their pharmacology, and these network relationships can be explored to discover new targets for established drugs.
The team ran a computer screen on 656 drugs that are already on the market, with known side effects and benefits, to predict which ones were most likely to bind to the 73 target proteins that appear on Novartis’ safety panel for testing drugs for side effects such as heart attacks. Then those targets were related to known side effects through a statistical method developed by NIBR. The computer model identified 1,241 possible side-effect targets for the 656 drugs. Of the total, 348 were already in Novartis’ database of drug interactions while another 151 hits revealed new side effects which were later confirmed in lab testing. They managed to predict potential side effects half of the time.
“The model isn't intended to replace existing safety approaches, like lab testing and animal models, but rather as a way to prioritize very early on what sorts of lab tests must be run or which research compounds might require fewer such tests than others,” said Michael Keiser, co-founder of SeaChange, whose work is focused on using computational simulations to identify new targets for known drugs.
Although it is not meant to replace testing, the new modelling has the potential to reduce animal testing. “Right now, some of the unsafe compounds that make it all the way to animal tests might be caught earlier by the computer model instead. This would reduce unnecessary animal suffering and also save money because each stage of the safety testing process is more expensive than the stage before it,” added Michael.
Adverse side effects are one of the pharmaceutical industry’s biggest sources of financial losses because, after inefficacy, they are the second most common reason that new drugs fail in clinical trials. The costs are huge, estimated at around US$1billion over a period of 15 years. With the computer model, it becomes possible to construct a preliminary “virtual safety panel” for almost any drugs or preclinical compound, said Michael. “You'd run your drug, or even your ten thousand compounds from which one or two might someday become drugs, and ask whether any undesirable side effects ‘light up’ against that panel,” he added.
Current scientific standards cannot predict how drugs will behave in relation to the targets linked to the side effects of medicines used clinically. Often, they hit more than 10 percent of the targets and the side effect targets are unrelated to previously known ones.
Can we expect a future when computer programs will be so sophisticated that they will be able to predict with complete accuracy the side effects of new drugs, making testing a thing of the past? “I don't know the answer to this question, though I certainly hope you're right," said Michael. "There are some exciting efforts going on right now in massive simulations of tissues, organs, and full systems. It's a tough problem; a grand challenge. Some of this effort is being supported by programs and agencies, such as in Europe. I'm excited to see where it goes.”
More details of the research appeared online this week in the journal Nature.