Researchers at the University of California, San Diego, have developed a computational method to study side effects caused by the widely prescribed selective estrogen receptor inhibitor Tamoxifen that they believe could be used to predict side effects for other drugs while they are still in development.
The method, developed by Philip Bourne of the university’s Skaggs School of Pharmacy and Pharmaceutical Services, is a virtual screening workflow that approaches protein-ligand docking from a different direction than most current methods.
Rather than screening thousands of small molecules against a single human protein to determine where it may bind, the UCSD team used the method, which integrates protein functional site similarity searching, small-molecule screening, and protein-ligand binding affinity analysis, to study the binding of a single drug molecule against a set of 825 “druggable” proteins from the Protein Data Bank.
In a validation study focusing on selective estrogen receptor modulators published this week in PLoS Computational Biology, they used the approach to identify a previously unknown protein target for the drug class.
“The identification of a secondary binding site with adverse effects opens the door to changing the drug to maintain binding to the intended target, but to reduce binding to the off-target,” UCSD said in a statement.
“What we found … is an explanation for a side effect of an existing drug,” Bourne told BioInform this week. “That is useful because now we can try to maintain” the binding in the intended target protein.
Bourne and his colleagues note in the paper that the approach could also be used “to explore off-targets binding for any existing pharmaceutical or compound of pharmaceutical interest for which a 3D structural model is available.”
Currently, they note, “we are beginning to systematically analyze all commercially available pharmaceuticals in an effort to explain any observed side effects.”
|
“At the end of the day, it’s one of a number of tools [that we can] … back up with experimental evidence … and the speed of the algorithm is also key.” |
Bourne said that one use the team is testing “is the notion that if you find another site where the drug binds … it then becomes a treatment for a different disease.”
He said that his group is currently in discussions with Novartis regarding ways the company could use the method to study the side effects of an undisclosed drug candidate. He added that he believes the method could have commercial potential.
Bourne said that the approach “would be useful as a precursor, to reduce the number of binding sites” in a drug-discovery setting.
Currently, he said, “that lead is tested biochemically, and then it’s tested in animals. But with this approach, you can potentially determine [that, for instance, a particular lead] would not likely be useful because it’s binding to a different receptor, which might have some fairly stringent side effects.”
Andrew Orry, a senior scientist with Molsoft, a virtual screening software firm based in La Jolla, Calif., said the UCSD method “is a good validation of the first step on the long road to computational prediction of ligand selectivity.”
While he noted that “methods have been around for a while for this kind of study,” including his company’s ICM software, “ it is good to see this method in practice with positive results.”
Still, Orry said he thinks many challenges lie ahead for the UCSD team if they intend to commercialize the approach because they must take into account pocket flexibility, crystal structure accuracy, and the ability to predict off-target sites in membrane proteins such as GPCRs and ion channels.
“In the future, this method could save the pharmaceutical industry time and money by flagging lead compounds that may have side effects and guide further experimental investigation into preventing them,” he said.
Even so, Bourne said he believes that the approach can enable researchers to use predicted off-targets to prioritize in vitro screening experiments, and that the off-target binding mechanisms will offer insights for optimizing drug leads.
“There’s nothing quite like what we’re doing, but there are a number of methods that people would argue about [as to] whether they are better or worse,” Bourne said.