The hope of developing better drugs

Bogyo, Pande discuss their research at chemistry meeting

New approaches to developing drugs were among the topics discussed by Stanford researchers last week during the 228th national meeting of the American Chemical Society in Philadelphia. More than 12,000 scientists attended. Following are summaries of two topics presented by Stanford scientists.

Cancer drug possibilities

Many cancer therapies take a “big stick” approach, targeting rapidly dividing cells in the body to stop malignancies in their tracks but often triggering horrible side effects in the process. New research by Matthew Bogyo, PhD, assistant professor of pathology, points toward the possibility of a different type of cancer drug – small molecules that would home in on the proteins tumors need without poisoning patients.

Previous studies have shown that an increase of certain proteins, called cathepsin cysteine proteases, is associated with tumor development. “The question is what are they doing and does it help to block them?” Bogyo said. To find out, he and his colleagues used fluorescent tags to track cathepsin activity in mouse models of two types of cancer. They found the cathepsins helped build blood vessels to the tumors as well as increasing tumor growth and invasiveness.

The team also tried giving a broad-spectrum cathepsin inhibitor to some cancerous mice and found that it slowed tumor development both in early and late stages of growth with no apparent side effects. The National Cancer Institute has since fast-tracked the inhibitor into studies as a potential drug. Meanwhile the research group is testing similar compounds to find those that bind specifically to “problem” cathepsins, instead of to all of them.

If one of the molecules turns out to vanquish tumors but not their hosts, it could become one of the first drugs of its kind, he said. “Most cancer drugs are very toxic,” he said. “There are very few rationally designed oncology drugs in use today that are made with a specific target in mind.”

Designing an inhibitor requires knowing which protein to inhibit and what the protein does. Finding out the latter is a particularly difficult task in the case of cathepsins because of the nature of proteases, Bogyo said. “Proteases take other proteins and degrade them into little bits, and so it’s hard to figure out what those proteins were doing,” he said. “We’re beginning by focusing on individual proteases, seeing what happens when you shut down a specific member of that family.”

The implications of finding the right inhibitor would reach far, Bogyo said, since cathepsins are associated with many different cancers and are involved in processes – such as blood vessel formation and tumor growth – critical to all tumors.

Harnessing the power of Pcs

Could your home computer help cure Alzheimer’s disease? The answer could be yes, according to Vijay Pande, PhD, assistant professor of chemistry and, by courtesy, of structural biology.

He’s devised a way to identify potential drug compounds by using a network of more than 150,000 home computers and some innovative algorithms. He said the method accurately predicts how well molecules will bind to a given protein. Proteins are the ubiquitous workhorses of living systems and most diseases can be traced to protein malfunctions of one kind or another, so designing a compound that binds to a particular protein is an early step in drug development.

“For almost 20 years, people have been talking about doing drug design computationally, but the real challenge has been getting sufficient accuracy,” Pande said. “Our goal was to come up with methods to push that accuracy to the point at which our methods are pharmaceutically useful.”

In the past, Pande said, computer predictions of binding strength between molecules and targeted proteins have been off by as much as 4 to 6 kilocalories per mol, rendering them essentially useless. But when he tested his new method, the results were accurate to within 1 kilocalorie per mol. “I think we’re at the point where pharmaceutical companies start to get interested,” he said.

To get those results, he tapped into Folding@Home, a global network of more than 150,000 home computers that run computations in the background, pooling their results to create computing power “greater than all the supercomputing centers combined,” in Pande’s words. He set up the network in 2000 to study protein folding and needed its power for this experiment because accurately predicting bonding energy requires “sampling” multiple conformations of a protein, a computationally demanding process.

Pande said this distributed-computing approach could be used to design new classes of antibiotics. And, as part of a current Folding@Home calculation on a protein critical to Alzheimer’s development, he hopes to identify molecules that would bind to the protein, pointing the way toward possible treatments.

Few researchers have a resource like Folding@Home at their fingertips, but Pande said his method could still have broad applications. The benefits of speeding up drug development could easily outweigh the cost of a multimillion-dollar supercomputer to a pharmaceutical company, he said. Also, several pharmaceutical companies are already harnessing the computers within their organization. As for academics, their time will come.

“One way to think of Folding@Home is as a time machine where we can do the sort of computational work now that would be very easy for any researcher to do in perhaps 10 years,” he said.