Computer Simulation of Protein Structure and Dynamics

Stanford University, Stanford CA

Investigators

Abstract

Research will focus on two areas of computational biology - (1)application of methods which explore conformation space for biomolecules, and (2) development of methods to help in protein structure determination both for soluble proteins and for membrane proteins. Understanding protein folding is important in a number of contexts such as protein design and the more general issue of finding general rules relating the folding process to the primary sequence of the protein. A major problem for computer simulation of protein folding is the fact that the dynamics takes place over many decades of time, from sub-picoseconds to milliseconds or even seconds. The approach we adopt is to constrain the motions of a protein so that it moves from an initial state, such as an unfolded state, to a final folded state in a fixed number of time steps. This "trajectory annealing" approach has the capability to explore a number of possible pathways by which a protein may fold while keeping the system at a physiological temperature. We will continue our present studies of trajectory annealing and apply them to two small proteins whose folding properties have been extensively sampled experimentally. The trajectory annealing simulations offer the possibility of interpreting those experiments in terms of changes in probability of various pathways for folding which are very difficult to observe experimentally. As a result of the tremendous progress in large-scale DNA sequencing projects, rapid growth in accumulation of biological sequence information has put strong pressure on the structural biology community to produce structural information for new genes with high throughput. Experimentally, large scale X-ray crystallography and NMR measurements now aim to determine all (1000 to 10000) available protein folds within a few decades or even years. However there remain many soluble proteins which have difficulties in crystallizing and for which NMR methods may be ineffective. Structural homology is a very powerful tool by which we can try to assign functions in silico to new genes which bear only remote if any association with known genes in terms of sequence homology. We will build on our current computational methods of ab initio structure prediction by adding additional physical information contained in small angle x-ray scattering (SAXS) data. This will improve our ability to find structural homologs of a given protein with proteins of known structure but for which there is little or no sequence homology. We will also extend our methods of reconstructing low resolution electron density maps from small-angle scattering data to x-ray scattering from membrane proteins bound in small vesicles. Some 25% of gene sequences in the database code for expression of membrane proteins. Although structures for several thousand soluble proteins have been determined, only a handful of structures for membrane proteins are known owing to the difficulties of crystallizing proteins which are naturally stabilized by the hydrophobic environment of the s urrounding lipids. By embedding a membrane protein in small lipid vesicles, we believe we can generate SAXS data which contains information about the shape and size of the protein in addition to data on the vesicle, and also about the x-ray interference pattern between the scattering from the vesicle and the scattering from the protein. By making a preparation in which vesicle sizes can be varied, we believe we can extend our present reconstruction methods to sort out the protein contributions to the scattering from the vesicle contributions and hence obtain low resolution structural information on the membrane protein.