U-M researchers seek to speed discovery of drugs


Researchers can invent and test millions of molecules quickly, but to develop successful new drugs, agrochemicals and other futuristic materials, they must first synthesize the molecules — and outcomes are a gamble.

To solve this problem, Timothy Cernak’s laboratory at the College of Pharmacy was awarded $3 million by Schmidt Futures recently to develop molecular reaction data using high-throughput experimentation, and the software to process it. The data will be available to all, and the software will be free to academia.

Cernak’s lab specializes in nanoscale synthesis, in which more than a thousand chemical reactions a day can be analyzed using high-throughput experimentation. They achieve this through miniaturization, the same way miniaturized transistors led to hand-held phones that would have been considered supercomputers a decade ago.

“High-throughput experimentation gives a systems-level look at any particular reaction. It lets you see the forest instead of the trees in a new chemical transformation,” said Cernak, assistant professor of medicinal chemistry in the College of Pharmacy and assistant professor of chemistry in LSA.

The project will yield 250,000 new chemical reactions to be made freely available for the development of machine intelligence in chemical synthesis. Cernak’s lab will also develop software called phactor, which will enable researchers to perform these experiments themselves. The software will be free for academic use and accessible to industry with a license.

“Schmidt Futures is excited about empowering researchers that are solving hard problems in science and society,” said Tom Kalil, chief innovation officer at Schmidt Futures. “By leveraging high-throughput experimentation, open datasets and machine learning, Dr. Cernak’s project could lead to the discovery of new classes of life-saving drugs.”

Current testing is slow and hinders development

Current chemical reaction testing methods are time-consuming and expensive because most chemical reactions fail, Cernak said. For this reason, drugmakers gravitate toward a few well-known chemical reactions. As a result, just five chemical reactions represent two-thirds of all reactions used in the pharmaceutical industry.

“Presently, these types of experiments are done with a handful of Excel worksheets and a whole bunch of Post-it notes,” Cernak said. “That’s fine for smaller experiments, but once you get up into a thousand reactions a day, a robust data management system is needed.”

Billions of unknown reactions exist, and predicting a recipe for each of them will be critical to the future of drug discovery, he said. While some reaction data exists in literature, it’s sparse and not machine readable.

Two of those popular reactions rely on a rare metal called palladium, and Russia is the world’s largest producer. Finding replacement metals will diversify supply chains and accelerate discovery of new bond types that will be important in medicines of the future, Cernak said.

Most nanoscale synthesis facilities are in industry, so a publicly available database and software is good news for consumers.

“The whole point of this project is to make data available to the masses,” Cernak said. “It’s likely that modern AI and machine learning could make short work of inventing the next generation of drug discovery reactions. There’s just no meaningful data available.”

Moonshot goal? Predict any chemical reaction

“In the future, we will design medicine based on patient needs and not on ease of synthesis,” Cernak said. “Today, we are somewhere in the middle — drug design necessarily weighs heavily on the ease of synthesis so we gravitate toward molecules that can be made by just a handful of reactions.

“Someday, I hope we can incorporate any of the billions of reactions that can be conceived today but not yet experimentally realized. We could make great strides in personalized medicine this way.”

Schmidt Futures is a philanthropic initiative founded by Eric and Wendy Schmidt.


Leave a comment

Commenting is closed for this article. Please read our comment guidelines for more information.