journal menu
![[Figure 6]](dz5234fig6mag.jpg)
|
|
Figure 6
Experimental assessment of refining a fully anisotropic model at various resolutions. The ADPs deposited for the 0.95 Å resolution refinement of 1lug are used as a gold standard. The top set of bars shows the extent to which ADPs obtained from refining an anisotropic model against data truncated to successively lower resolution are a better approximation to the corresponding `true' ADPs than would be obtained from a purely isotropic model. This comparison uses the statistic SUV (Merritt, 1999b  ). The green portion of each bar corresponds to atoms with SUV > 1 + ∊, indicating that the electron density described by anisotropic treatment correlates better with that of the reference model than the density described by isotropic treatment. The red portion of each bar corresponds to atoms with SUV < 1 − ∊, indicating that anisotropic treatment yields a worse approximation to the reference density distribution than isotropic treatment. Values of SUV near 1.0 indicate that the anisotropic and isotropic models for that atom are equally good (or poor) approximations to the reference ellipsoid. The yellow portion of the bar is drawn for ∊ = 0.01. The lower set of bars show the extent to which the anisotropic ADPs obtained from refinement at truncated resolution diverge from those in the 0.95 Å reference model. If the model ADP U and the reference ADP V are identical, then the Kullback–Leibler divergence KLUV for that atom is equal to zero. Larger values of KLUV indicate increasing disparity between the electron-density distributions described by U and V. The height of each bar indicates the median value of KLUV calculated for all 2120 protein atoms in the refinement at that resolution. The rightmost bars show the same statistical assessments applied to a hybrid model containing 16 TLS groups in addition to a single parameter Biso for each atom. The quality of the refined hybrid model is only weakly sensitive to truncation of the data in this resolution range; the bars shown are for refinement against data truncated to 1.5 Å resolution. |
![[Figure 6]](dz5234fig6.jpg)
 | STRUCTURAL BIOLOGY |
ISSN: 2059-7983
access
