Computing for Cancer Research

[编者的话]

Screensaver program allows for computational chemistry research on "virtual supercomputer

By Larry Hand

How do you "rapidly" screen 250 million small molecules--eventually 1 billion small molecules--for their cancer-fighting potential? Enlist, literally, a "world" of computers to perform the computational chemistry.

That's the goal of Graham Richards, chairman of the chemistry department at Oxford University and director of the National Foundation for Cancer Research Center for Computational Drug Design, a virtual collaboration set up last August with a $750,000 NFCR grant.

Richards recently got some help from a computer program called THINK, developed by United Devices of Austin, Texas, and from another company that knows a little about computers: Intel Corp. in Santa Clara, Calif. In early April, the chipmaker launched the Intel® Philanthropic Peer-to-Peer Program in a special news briefing at company headquarters. Joining Intel president and CEO Craig Barrett on the podium were Richards; Sujuan Ba, science director for the NFCR; Ed Hubbard, CEO of United Devices; and John Seffrin, president of the American Cancer Society.

Through this effort, all of those involved hope that eventually millions of computer users will each perform their own form of cancer research by allowing THINK to operate in the background or as screensavers on their computers, and communicating results of computations to a central server when connected to the Internet. Computer users can download THINK from the Intel (www.intel.com/cure) or from United Devices (members.ud.com/download/gold).

"What we're talking about is the possibility of getting something like 50 teraflops of computer capacity [trillions of operations per second]--for almost free," explains Barrett, by providing the ability for people to donate idle time on their computer when attached to the Internet. As of April 27, it looked as though the project was well on its way with 318,612 "members" participating on 443,579 computers that had spent 39,834,919 hours of processing time on THINK.

By providing a "virtual supercomputer" to the Oxford scientists, "we hope to cut the time it takes to bring new cancer treatments to patients by three to five years," maintains Ba of NFCR. "This worldwide THINK program ... will identify approximately 10,000 new drug candidate molecules for further biological testing, and we are hopeful that one or several breakthrough therapeutic agents will be identified to cure leukemia," she adds.

 "Although there are only tens of thousands of genes [and] a few hundred thousand proteins, [of] the little molecules that interact with those proteins, there are essentially astronomical numbers--possibly 18 billion," says Richards. For this project, he adds, "we have a database º[with] 250 million small molecules in it, orders of magnitude more than all the pharmaceutical companies in the world have. We will create an even bigger database; we would like to get to a billion small molecules, and we want to see how each of these little molecules will bind to and be a potential threat to attack [the target]."

Richards' first protein target is superoxide dismutase. "In leukemia cells, there is an excess of these superoxide dismutase molecules. Recent fundamental biological work suggested that if you could block that enzyme, you could damage the leukemia cell,"1 Richards explains. Other targets will be the ras protein and vascular endothelial growth factor. The THINK program scans the small molecules and scores "hits" when binding sites are identified. "Out of this we will get an indication of what small molecules are worth starting out along the line of creating new drugs," he adds.

The THINK program displays the molecules in three dimensions as they move about on screen and counts the hits as they occur. The program downloads the molecular coordinates when the computer is connected to the Internet, then uploads the computational results usually about a day later, downloads another set of molecular data, and another cycle begins.

Richards expects to get hundreds of thousands of hits from this project. Then analysis--largely done in the same manner--will be conducted to determine the most likely possibilities for developing drugs. Richards and the NFCR Center will own the rights to the data collected from the project, which will be used in nonprofit ways to develop drugs.

Planning on the project, involving the Oxford center, the NFCR, and United Devices, began last September, and a test database had been running about three months prior to the April announcement.

"Genomics research and combinatorial chemistry have provided scientists with tremendous amounts of genetic and chemical information [that] we need to fight this terrible disease," Ba says, "but scientists continue to struggle with ways to identify targets and deliver safe and effective drugs to patients. Currently, it takes about 10 to 12 years to put a drug on the market. ... Now, with the combination of efficient computational drug design and advances in computer power, we are entering a new era in cancer research."

Richards, who in 1998 earned the Mullard Award from the Royal Society for developing computational methods for molecular design, sums up the project this way: "I think we're having the opportunity here to turn a screensaver into a life saver, which is actually much better than flying toasters."

Larry Hand can be contacted at lhand@the-scientist.com.

1. P. Huang et al., "Superoxide dismutase as a target for the selective killing of cancer cells," Nature, 407:390-5, Sept. 21, 2000.