Summary Knowledge-Based B-Factor Restraints for Results

Example Refinement

To demonstrate the effect of the proposed restraints a pair of refinements were performed with the same starting coordinate file. The structure was a complex of thermolysin with the inhibitor phosphoramidon [Weaver et al., 1977]. Even though the crystals diffracted well the data were only collected to 2.3Å resolution. The starting model was constructed by placing the unrefined inhibitor model[Weaver et al., 1977] into the model of inhibitor-free thermolysin described previously [Holmes & Matthews, 1982]. The thermolysin model had been refined with PROLSQ[Hendrickson & Konnert, 1980] and the B-factors restrained in the usual fashion for that program.

In the first refinement this model was refined for 30 cycles using the TNT refinement package[Tronrud et al., 1987] without application of any stereochemical restraints on the B-factors. The final model had an overall B-factor discrepancy (based on the new library) of 14.0Å . In the second refinement the new library was enforced and the final model had a root-mean-square B-factor discrepancy of 6.4Å .

Even though the agreement to the restraints was much improved the R value rose only 0.1% (from 13.6% to 13.7%). This result tends to confirm that the restraints add information which is not present in diffraction data of 2.3Å resolution.