W0116

Eigensystem Analysis of Crystallographic Refinement at Medium and High Resolution. Lynn F. Ten Eyck and Kevin Cowtan, San Diego Supercomputer Center, University of California at San Diego, La Jolla, CA 92093-0505 USA

Analysis of the eigenvalues and eigenvectors of the refinement equations directly answers long-standing questions concerning refinement methodology. Effects of parameter restraints and constraints on the final results, effects of different atomic displacement parameter models, and effects of choices of data cutoffs based on resolution and/or precision are explicitly revealed by these techniques. Conventional wisdom and rules of thumb are given quantitative interpretations which reflect the peculiarities of specific datasets.

The power of this approach is demonstrated by studies on a ferredoxin for which separate datasets and refinements are available at 1.9 Angstroms (room temperature) and 1.3 Angstroms (low temperature). The eigenvalue and eigenvector analysis identified subtle problems in the structure, and clearly demonstrated the precision with which the various parameters are determined (or not). This analysis also clearly demonstrates the need for 64-bit arithmetic in performing these calculations.