Protein structure refinement through effective sampling and scoring. Detailed structural information is essential in understanding biological processes in detail and in allowing the rational development of therapeutic strategies against a variety of diseases. Experimental methods allow the accurate determination of high-resolution structures, but are encumbered by significant effort and experimental constraints. As an alternative, computational methods can predict protein structures to some degree of accuracy. However, it has remained a challenge to routinely predict protein structures at near-experimental accuracy. A high level of resolution may be reached through refinement of initial models. Successful protein structure refinement requires sampling methods that can generate native-like conformations and scoring methods that are able to identify the most native structures from a set of candidates without any knowledge of the true experimental structure. In order to achieve these goals novel protein structure prediction and refinement protocols are developed. In particular, effective conformational sampling strategies based on existing and new methods with constraints to reduce conformational search space are introduced;novel statistical methods to enhance and combine existing scoring functions in the selection of refined models from a set of decoys are developed;and an intermediate resolution model PRIMO is developed to obtain a better balance between energetic accuracy, model resolution, and sampling efficiency. These new methods are combined into an integrated refinement strategy and applied in the context of an automated protein structure pipeline.
New computational methods for the accurate prediction of protein structures are developed as an alternative to experimental approaches. Such structural information is crucial in understanding detailed biological mechanisms and allowing the development of therapeutic strategies against a variety of diseases.
Showing the most recent 10 out of 39 publications
