It’s the drug discovery equivalent of looking for a book on Amazon versus at your local library. Researchers have scanned a chemical database containing some 170 million molecules—100 times larger than previous databases—to identify a handful of new compounds that could serve as starting points for novel antibiotics and antipsychotic medications. The resource is expected to grow to more than 1 billion molecules over the next year, making the technique increasingly powerful as time goes by.
“For the [drug discovery] industry as a whole, this is a great thing,” says Jeff Blaney, who directs computational chemistry research at the biotech company Genentech in South San Francisco, California, and was not involved in the work. Sampling a huge variety of molecules against particular disease targets means higher chances that one will prove to be a successful starting point for drug discovery, he says. “More shots on goal will help.”
To scan possible drug molecules, researchers use “virtual screening.” The approach evaluates how well a potential molecule might bind to a protein or other biological target in the body. Scientists use software called molecular docking programs, among others, to probe thousands of orientations a molecule might take as it binds its target. The binders are then ranked, and the tightest ones synthesized so they can be tested experimentally.
The problem is that the number of possible druglike molecules—1063—is impossibly vast, similar to the number of atoms in the universe. Rather than focus on the majority of those molecules that will likely never be made, researchers have begun to team up with chemical supply companies that can make vast libraries of compounds on demand. One such company, Enamine, in Kyiv, for example, starts with 70,000 small chemical building blocks that they can connect to one another using 130 well-known chemical reactions. That’s allowed the company to assemble a database of more than 700 million compounds that it can make to order in small amounts—a library that’s already about 100 times the size of most libraries scanned by pharmaceutical companies.
In 2016, researchers led by Brian Shoichet, a computational chemist at the University of California, San Francisco, scanned Enamine’s database, which at the time included 3 million molecules. They pinpointed a potential opioid painkiller that likely would lack the addictive properties of today’s opioid drugs. A biotech company called Epiodyne is now working to turn this lead into a medicine.
Now, Shoichet and his colleagues have screened 170 million of Enamine’s compounds against two targets: a focal point of some antibiotics, known as AmpC β-lactamase, and the D4 dopamine receptor protein, a target for antipsychotic medications. Because of the sheer size of the database, Shoichet says, “We were terrified as to how we were going to find the interesting molecules, spotting the signal in the noise.”
So, the researchers decided to first test whether their software could spot the hundreds of already known inhibitors of these targets amid a library of 170 million molecules. Using a cluster of 2000 computer processors, they found that the top-scoring molecules included the known inhibitors and their structural cousins. Then, Enamine scientists synthesized hundreds of high-scoring compounds that hadn’t previously been identified. Of these, 24% were found to bind tightly with the D4 receptor and 11% against AmpC β-lactamase, far higher hit rates than other virtual screening programs, the researchers report this week in Nature.
The higher hit rates increase the odds that one will eventually lead to a medicine, Shoichet explains. The performance is likely due to the fact that the large database is sampling families of chemical structures that have never been scanned before. “I feel like a door has finally popped open,” for the field of virtual screening, he says.
In addition to aiding drug discovery, the ever-growing database will also help myriad basic biology researchers, says Laurie Nadler, a program officer for the National Institute of Mental Health in Bethesda, Maryland, which helped sponsor the research. As researchers discover novel protein targets in the body related to different diseases, they’ll be able to scan the publicly available database for compounds that could hit those targets, she says. “The large size of the virtual library and the fact it is publicly available will have a large impact on pharmacology and drug discovery.”