The University received $5 million from the National Institutes of Health (NIH) to develop the world’s leading resource of high-quality experimental data sets of drug-making compounds that will ultimately take computer-aided drug design to a new level.
The resource will house the data needed to improve computer programs that can predict the effectiveness of potential new drugs, says Heather Carlson, associate professor in the College of Pharmacy and Medicinal Chemistry.
“What we intend to do is make the best data available so that people can use computers to do better drug design,” says Carlson, who will oversee the creation and operation of the database, called Community Structure-Activity Resource (CSAR). “Computers do a good job but there is really room for improvement.”
The data for drug compounds is spread among pharmaceutical companies and academic labs, and CSAR will combine existing and older data from those sources and generate new data. Carlson hopes to have the internal data online and accessible by year’s end and to start accepting data deposits shortly after. The Web-based resource will freely be available to scientists and others interested in this information.
“We’re talking about at least a hundred companies, each with data on thousands of compounds,” Carlson says. “Access to that kind of data really changes the landscape.”
Collaboration of three U-M centers — one of which houses many former Pfizer employees — allowed the University to land the award, Carlson says.
The project is also heavily supported by the Center for Structural Biology in the Life Sciences Institute (which also includes former Pfizer scientists) as well as scientists from the College of Pharmacy, LSI and the Medical School.
The database is important because it will include detailed molecular information about proteins that bind small, drug-like molecules called ligands. Most drugs work by latching onto proteins and altering a biological process, turning it on or off. Researchers can use computers to study the structural and biophysical properties of a target protein and, from among tens of thousands of possible ligands, predict those that bind to the protein in a potentially useful way. In this way, they identify ligands that may eventually become lead compounds for drug discovery or indicate which compounds may interact with other proteins and cause unwanted side effects.
“The ability to screen compounds and accurately predict their binding properties using only computers would greatly impact the drug development process and many other aspects of biomedical research,” says Jeremy Berg, director of the National Institute of General Medical Sciences of the NIH. “This resource has been established to make important structural and binding data available so researchers can tackle this problem.”
