The pharmaceutical industry dreams of using computers to predict which potential drugs will work best and have the fewest side effects. Until recently, the software that performs such simulations has been slow, difficult for the average chemist to access and use, and worst of all, often inaccurate; but a collaboration between University of North Carolina at Chapel Hill computational chemist Alex Tropsha and IBM is making predictive drug screening easier and more accessible. Researchers who have encountered difficulty with a drug candidate can submit data on its structure and biological functionality to Tropsha's program via a Linux-based Web portal designed by IBM. Running on a cluster of high-powered computers, Tropsha's software screens a large database of compounds to help identify more-promising candidates. "Now, all a chemist who uses the program has to know is chemistry," says Tropsha. The program has already singled out compounds that could help treat conditions such as convulsions and hemophilia, and Tropsha hopes to commercialize the software this year.