One of the main problems in computational protein folding is to devise methods that can go beyond helix formation, and that can assemble secondary structures into tertiary conformations. For example, chain representations using limited phi/psi choices can not handle secondary structure assembly (SSA). In native protein structures, helices and strands that are not sequential neighbors can come close in space and pack tightly against one another. But in the representations the packing positions imposed by the limited orientation choices will result in either steric conflicts or suboptimal packing. We are developing a new search approach for tertiary assembly. Guided by standard templates of tertiary structure packing, taken from the PDB, and standardized representations of helices and strands, we extend an algorithm called Geocore, to search for good tertiary packing. To enable two such idealized secondary structures to pack, the phi/psi's of intervening loop residues are adjusted as continuous functions. In this search, in addition to the search performed in Geocore, segments of a sequence are designated as helices or strands and the secondary structure assembly is constructed by adding one secondary structure at a time. We have tested this strategy on proteins up to chain lengths of around 100 amino acids. Comparison of the steric packing predicted by the algorithm with the known native structures indicates that this strategy is promising.
Showing the most recent 10 out of 508 publications
