The amino acid sequence of a protein uniquely determines its tertiary structure. Deciphering this relationship, the protein folding problem, has become increasingly important to molecular biologists. DNA sequencing has become routine, but structural experiments remain very difficult. Computational strategies are needed to help address this problem. This proposal describes a strategy to identify the location of alpha-helices and beta-strands throughout the sequence. A rationale is offered for employing neural networks and pattern based algorithms to address the secondary structure prediction problem. Once secondary structure is located, computational methods exist for generating plausible tertiary structures. However, these combinatorial strategies give rise to a large number of alternative structures which are difficult to distinguish from the correct fold. Simplified potential functions are proposed as a method for overcoming this structure evaluation problem. The properties of a non-lattice based simplified representation of a polypeptide chain will be explored to aid in the construction of an appropriate simplified potential function. Collaborative ventures are planned to experimentally test the merits of existing algorithms for predicting protein structure. In collaboration with Dr. Bunn at Harvard, the relationship of the erythropoietin sequence to its structure and function will be explored. In collaboration with Dr. Wang at UCSF, the merits of a proposed structure of hypoxanthine guanine phosphoribosyl transferase will be studied using site directed mutagenesis. An exploration of the possibility of grafting the active site of one enzyme onto the structural scaffold provided by another protein will be studied in collaboration with Dr. Craik at UCSF and Dr. Wells at Genentech.
Showing the most recent 10 out of 26 publications