Protein motion and conformational change plays a critical role in vital biological functions such as immune protection, enzymatic catalysis, and transport of metabolites. The study of protein motion provides an important link between structure and function, and enhances our understanding of ligand-protein interactions and our ability to discover new drug ligands. With the advances in X-ray crystallography and nuclear magnetic resonance spectroscopy, an increasing number of structures have been determined for proteins in different conformations. Since the data is readily accessible in the Protein Data Bank (PDB), it is easy to use superposition to demonstrate that conformational change has occurred. It is less easy to see, by comparing two structures, exactly what type of motion has enabled the conformational change. An important issue that has not been adequately addressed is how to distinguish genuine structural change from random noise.

We present an efficient multi-scale method that deals with noise in a principled manner. We develop simple models of the error in determining atom positions for a protein at a chosen scale and how this error propagates through the refinement process. We use these to estimate a threshold that distinguishes conformational change at that scale. By evaluating a hierarchy of scales, we can identify the location and extent of the change. Our ultimate aim is a flexible model of the protein, which consists of fixed bond lengths, fixed bond angles, and some variable torsional angles. This flexible model can be used to compute plausible motions that are consistent with experimental data.