The right molecule can make a fortune by treating a disease or creating a new property in a material. Companies use techniques such as nuclear magnetic-resonance imaging to gather information about the structures of thousands of promising molecules. Often, though, these data languish in filing systems, so companies can essentially "lose" information on the one molecule in 10,000 that might lead to a breakthrough.
Help may be on the way in the form of a new computer-based system, developed by scientists at the University of Georgia's Complex Carbohydrate Research Center (CCRC), that can quickly identify molecules based on the data generated by many types of analytical instruments. So-called artificial neural networks, which I.D. molecules by matching patterns in the data and by "learning" through experience to match more precisely, make the new system work. The university has spun off ANN Technology Inc. to commercialize the technology. If ANN secures further financing, an entry-level system is expected on the market within six months, says Peter Albersheim, co-director of the CCRC.