Evaluating experimental and theoretical measures of protein conformational dynamics

Abstract

Molecular biologists have traditionally interpreted the B-factor data of a protein crystal structure as a reflection of the protein's conformational flexibility. Crystallographers, in contrast, are wary of assigning too much significance to B-factors since they can also be attributed to processes unrelated to conformational dynamics such as experimental imprecision; crystal imperfections; or rigid body motion. In this study, the usefulness of both isotropic and anisotropic B-factors as measures of conformational dynamics were evaluated using high resolution structures. Alpha-carbon B-factor values were analysed in relation to structural properties generally accepted to be correlates of conformational variability. The influence of secondary structure, amino acid type, surface exposure, distance to the centre of mass and packing density were investigated. The results support the argument that B-factors measure conformational variability by demonstrating that atoms with the highest B-factors are typically located in regions expected to have a high degree of conformational freedom. Nevertheless, the results also highlight some of the limitations of crystallographic data. Despite using high quality crystal structures, only very general qualitative trends between B-factors values and the properties investigated could be established. Thus, B-factors appear to be influenced, to a significant degree, by the numerous sources of error in a crystallographic experiment. By considering proteins with multiple published crystal structures, the existence of consensus B-factor profiles were identified. These consensus profiles were hypothesised to represent the dynamics within the crystal with a high degree of accuracy since much of the variation between individual experiments would be eliminated. However, when compared against measurements derived from molecular dynamic simulations, these consensus profiles only weakly correlated with the predictions of the computer models. Therefore, although there is some evidence to suggest that B-factors reflect conformational variability, B-factors cannot be assumed to be reliable descriptors of the internal dynamics of a protein within a crystal.