Application of known X-ray phases in the crystallographic study of a small protein.

Phase information, assumed known from three-beam diffraction experiments, has been used successfully as input to direct methods to re-determine the crystal structure of rubredoxin. With data at 1.54 A resolution, starting sets containing 26 single phases, or alternatively 45 triplet phases, in both cases known with a mean error of +/-22.5 degrees, were sufficient to initiate solution of the structure. Conventional figures of merit were employed in the early stages to reject the majority of the incorrect phase models. The presence of a FeS(4) cluster in the structure was used in the interpretation of the initial maps. Phase sets including 2500 E's with a mean single phase error <Deltaphi> approximately 70 degrees or a mean triplet phase error <DeltaPhi(3)> approximately 80 degrees, both relative to the model from the crystallographic refinement, could be refined and expanded by Fourier recycling using the SAYTAN formalism. Several parameters have been varied to study their influence on phase expansion and refinement.