Number 73 (Story #2), March 27, 1992 by Phillip F. Schewe and Ben Stein A QUICK METHOD FOR PREDICTING HOW MANY PROTEINS FOLD has been developed by researchers at the University of Illinois. It is a relatively straightforward matter to learn the chemical sequences of unfolded proteins, but it takes months with conventional x-ray and NMR techniques to determine a single protein's final three-dimensional shape. At the March Meeting, Peter G. Wolynes (217-333-7385) and his colleagues described a model which allows researchers to take proteins whose shapes are known and predict the shapes of other proteins---having similarities in chemical sequence of as little as 17 percent---using about an hour of supercomputer time. Since a protein's transition from its original unfolded state to its final three-dimensional shape is complicated, with many intermediate states arising from complex interactions between molecules, the Illinois team devised an algorithm in which certain energy states in the protein are favored over others, based on the characteristics of proteins with known shape. According to Wolynes, the predicted shapes can be tested using the conventional methods.