Chapter 5

Abstract:

We present a method for predicting the complete conformation of a protein from its C coordinates based on the Probability Grid Monte Carlo (PGMC) Method described in Chapter 4. Unlike most methods designed to solve this problem, the PGMC Method does not attempt to fit known polypeptide conformations onto the C framework. Rather, conformational propensities for individual residues are used to guide conformational searches while the protein is built from the amino-terminus to the carboxy-terminus. Therefore, no structural homology to other known structures is required. We present results for a number of proteins and show that both the backbone and sidechain can be accurately modeled using the PGMC method. Backbone atoms can generally be predicted to within 0.6 Å of their X-ray crystal structure coordinates, while the total rms for all atoms can be predicted to 1.7 Å or better. The method is also used to build all-atom protein models from C coordinates derived from lattice-based methods of protein structure prediction, through the use of a ``C Forcefield.''