This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. The subproject and investigator (PI) may have received primary funding from another NIH source, and thus could be represented in other CRISP entries. The institution listed is for the Center, which is not necessarily the institution for the investigator. We developed a computational approach to describe protein folding using a reduced model that contains three major features, a beta carbon side chain representation, backbone dihedral rotations as the only allowed motions, and implicit solvent. The model captures the underlying all-atom information by using two strategies. The first consists in constructing a reduced statistical potential defined at the level of the backbone atoms plus beta carbon or centroid atoms representing the side chains. The second strategy involves sampling conformations with a backbone rotamer library that is built from the PDB by combining joint information about amino acid sequence and Ramachandran basin (RB) occupancies. Protein structures are obtained by minimizing the reduced statistical potential with a simulated annealing algorithm that samples backbone conformations from the library, while constraining them to remain within the native Ramachandran basins. We plan to evaluate the performance of the algorithm by running simulations on an heterogeneous set of 90 globular proteins between 60 and 100 residues long. Each simulation takes an average of 24 hours of CPU time on a standard PC (Pentium 4, 2.8Ghz). The algorithm has not been parallelized, so each simulation would run independently on each node of the assigned cluster. The code was been written in standard C++ and tested on a variety of x86 Linux distributions. Furthermore, we are also interested in generating simulations with different reduced statistical potentials. This would account of a total number of about 300 simulations. This computational experiment would provide a crucial test of the algorithm in regards to its ability of predicting tertiary packing given local backbone constraints and the use of a greatly reduced representation.
Showing the most recent 10 out of 292 publications
