Time-averaging crystallographic refinement: possibilities and limitations using α-cyclodextrin as a test system

Abstract

The method of time-averaging crystallographic refinement is assessed using a small molecule, a-cyclodextrin, as a test system. A total of 16 refinements are performed on simulated data. Three resolution ranges of the data are used, the memory relaxation time of the averaging is varied, and several overall temperature factors are used. The most critical factor in the reliable application of timeaveraging is the resolution of the data. The ratio of data to molecular degrees of freedom should be large enough to avoid overfitting of the data by the timeaveraging procedure. The use of a free R factor can aid in determining whether time-averaging can be reliably applied. Good ensembles of structures are obtained using data up to 1.0 or 2.0 A resolution. Comparison of electron-density maps from timeaveraging refinement and anisotropic temperaturefactor refinement indicates that the former technique yields a better representation of the exact data than the latter.

Fluctuation and Cross-correlation Analysis of Protein Motions Observed in Nanosecond Molecular Dynamics Simulations

“…Using linear regression for residues 5 to 55, we found the following correlation coefficient values: 0.35 (4PTI versus 5PTI); 0.22 (4PTI versus 6PTI); and 0.64 (5PTI versus 6PTI). Recently, it has been shown by comparison of electron-density maps obtained from anisotropic B-factor refinement and from time-averaging MD refinement using exact data that the former technique yields a worse representation of the exact data than the latter (Schiffer et al, 1995). The influence of the force field (reason (4)) is difficult to assess but should not be underestimated.…”

Modelling dynamics in protein crystal structures by ensemble refinement

“…This method introduced by Torda et al (1989) and implemented in macromolecular crystallography by Gros et al (1990), showed a reduction in R -value. However, cross-validation introduced subsequently (Brünger, 1992) revealed chronically over-fitted models (Burling and Brunger, 1994; Clarage and Phillips, 1994; Schiffer et al, 1995). …”

Thermostat Algorithms for Molecular Dynamics Simulations

“…Experimentally derived boundary conditions are used to explicitly enforce agreement between a simulation and some experimental result. These may be applied to enforce, e.g., the reproduction of (average) electron density maps from X-ray crystallography [27,28,29,30], or the agreement with (average) interatomic distances and J-coupling constants from NMR measurements [31,32,33,30].…”