Structural quality assurance.

The experimentally determined three-dimensional (3D) structures of proteins and nucleic acids represent the knowledge base from which so much understanding of biological processes has been derived over the last three decades of the twentieth century. Individual structures have provided explanations of specific biochemical functions and mechanisms, while comparisons of structures have given insights into general principles governing these complex molecules, the interactions they make, and their biological roles. The 3D structures form the foundation of structural bioinformatics; all structural analyses depend on them and would be impossible without them. Therefore, it is crucial to bear in mind two important truths about these structures, both of which result from the fact that they have been determined experimentally. The first is that the result of any experiment is merely a model that aims to give as good an explanation for the experimental data as possible. The term structure is commonly used, but you should realize that this should be correctly read as model. As such the model may be an accurate and meaningful representation of the molecule, or it may be a poor one. The quality of the data and the care with which the experiment has been performed will determine which it is. Independently performed experiments can arrive at very similar models of the same molecule; this suggests that both are accurate representations, that they are good models. The second important truth is that any experiment, however carefully performed, will have errors associated with it. These errors come in two distinct varieties: systematic and random. Systematic errors relate to the accuracy of the model—how well it corresponds to the true structure of the molecule in question. These often include errors of interpretation. In X-ray crystallography, for example, the molecule(s) need to be fitted to the electron density computed from the diffraction data. If the data are poor and the quality of the electron density map is low, it can be difficult to find the