How much experimental data is needed to predict a protein's structure?

COMP 353

Christina R. Crecca, crecca@qtp.ufl.edu and Adrian E Roitberg. Department of Chemistry, University of Florida, NPB 2331, PO Box 118435, Gainesville, FL 32611-8435
Protein structure determination is crucial in the elucidation of protein function. Present experimental methods, however, are costly and time-consuming. But how much experimental data is really needed to predict a protein's structure? We will present a fast and inexpensive structure prediction method that seeks to answer this question. Our method involves three steps: (1) building a decoy set, (2) using a small number of experimental inter-residue distances, and (3) comparing the measured distances with those calculated for each decoy. We postulate that structures with several similar distances will also have similar three-dimensional structure. Using two different types of decoy sets, general and specific, we make structure predictions for several CASP7 target proteins. We will also compare our method to others.