Statistical modelling and likelihood-based choice in macromolecular crystallography

Statistical modeling allows considering the structure under study as an element of an ensemble of structures and assigning common properties of the ensemble structures to a particular structure. Such common properties may be derived sometimes more easily for the whole ensemble than for the particular structure separately. The utility of the properties derived depends on how adequately the ensemble is introduced. The statistical likelihood may be used as the basis to choose the most reasonable statistical model. Several examples of the use of this approach in studies of structures of biological macromolecules by X-ray diffraction methods are discussed.