On the Use of Partial Structure Information in Ab Initio Direct Methods

Abstract

Traditionally, the phrase ‘ab initio direct methods’ has been reserved for processes that intend to derive a crystal structure model from the observed structure factor (s.f.) magnitudes only with as little structural information as possible, often the contents of the unit cell only. This expresses the underlying wish to solve the phase problem exactly in one step only or, as a next best solution, in as few steps as possible. In small-molecule crystallography, with its high resolution data and a large ratio of data/variables, this goal may not seem unrealistic because only a small amount of structural conformations will fit the observed data. In contrast, data of macromolecular compounds are often of lower resolution. Thus, together with the much smaller ratio of data/variables, a much larger ensemble of conformations is feasible, even conformations that differ appreciably. Therefore, to ‘solve’ a macromolecular crystal structure, i.e to select the most likely conformation(s) given the observed data, any knowledge should be exploited that can limit the conformation(s) to be considered. The initial macromolecular model often consists of a molecular envelope at low resolution, e.g. obtained from (multiple) isomorphous replacement, anomalous scattering or molecular replacement data.