Merritt Ethan A
Department of Biochemistry, University of Washington, Seattle, WA 98195-7742, USA.
Acta Crystallogr A. 2011 Nov;67(Pt 6):512-6. doi: 10.1107/S0108767311034350. Epub 2011 Oct 13.
Crystallographic structural models for macromolecules have typically included an isotropic displacement parameter B(iso) for each atom. In cases where the structural model instead includes anisotropic displacement parameters U(ij), the derived quantity B(eq) can be substituted for B(iso) for many purposes. B(eq) is not, however, the best predictor of the value B(iso) that would hypothetically have been obtained by direct refinement of an isotropic model. A new entity B(est) is proposed that represents an estimate for B(iso) that minimizes the Kullback-Leibler divergence from a paired anisotropic model. In general B(est)/B(eq) < 1, with the difference between the two values becoming larger for atoms that are more anisotropic. Although this difference does not affect direct refinement of either isotropic or anisotropic models, it is relevant to any analysis that compares isotropic and anisotropic models of the same underlying structure. In particular, it may lead to improved selection of multi-group TLS models based on analysis of an initial isotropic refinement.
大分子的晶体学结构模型通常为每个原子包含一个各向同性位移参数B(iso)。在结构模型包含各向异性位移参数U(ij)的情况下,导出量B(eq)在许多情况下可替代B(iso)。然而,B(eq)并非假设通过直接精修各向同性模型所得到的B(iso)值的最佳预测指标。现提出一个新的实体B(est),它代表对B(iso)的估计,能使与配对各向异性模型的库尔贝克-莱布勒散度最小化。一般来说,B(est)/B(eq) < 1,对于各向异性更强的原子,这两个值之间的差异会更大。虽然这种差异不影响各向同性或各向异性模型的直接精修,但对于比较相同基础结构的各向同性和各向异性模型的任何分析都很重要。特别是,基于对初始各向同性精修的分析,它可能会改进多组TLS模型的选择。