The accuracy of refined protein structures: comparison of two independently refined models of bovine trypsin
The accuracy of refined protein structures: comparison of two independently refined models of bovine trypsin
复制标题
精制蛋白质结构的准确性：两种独立精制的牛胰蛋白酶模型的比较
DOI：
--
复制
发表时间：
1978
期刊：
影响因子：
--
通讯作者：
R. Stroud
中科院分区：
文献类型：
--
作者：
J. L. Chambers;R. Stroud
二异丙基氟磷酸盐抑制的牛胰蛋白酶的结构已被细化到标准晶体学残差R = 0.157，在1.5/~，分辨率为约束模型（C&S坐标）。苯甲脒抑制的牛胰蛋白酶也已经由Bode & Schwager独立地精制（在1.8yL/k下R = 0.229）[J. Mol. Biol.（1975），98，693-717]（B&S坐标组）。在对不同抑制剂进行适当校正后比较这些结构以及随后的结构差异，允许实验测定结构差异，这对精确坐标组中的误差设置了上限。在确定的区域，模型非常相似。内部主链原子之间的平均位置差为0.146 A（0.163 A r.m.s.）;然而，存在一些大至3-9 A的差异，主要与不确定的外部侧链有关。偏差的大小强烈依赖于比较的结构区域，并且与我们分析中的精细热参数密切相关。© 1979 International Union of Crystallography 1862牛胰蛋白酶的两个独立改进模型C&S原子位置的估计误差总体平均为0.15-0-20 /k，范围从确定良好的区域的约0 -10/k到确定不良的区域的1 A或更大。预计B&S头寸的误差会更大一些。从B&S坐标[F~ss）]计算的结构因子的比较给出了在1.8/k处R = 0.238的C&S观测数据的残差（非常接近它们自己的残差0.229），这表明来自两个“真实的”结构之间的真实的差异或来自任一观测数据集中的误差的贡献与精确坐标集中的误差相比是小的。在1.5A处，从C&S模型计算的Fe~Bs）和结构因子之间的残差为0.253，这表明两个模型在系统上相同的方式下与“真实”结构没有区别，并且进一步的细化应该改进任一模型。0.253值与1.8 A处的0.229残差值的相似性也表明，C&S和B&S模型之间的差异与B&S结构和假设的“真实”结构之间的差异一样大。因此，两个模型之间的差异给出了对具有类似于Bode & Schwager的细化统计数据的结构中可能出现的错误类型的合理估计。
The structure of diisopropyl-fluorophosphate-inhibited bovine trypsin has been refined to a standard crystallographic residual of R = 0.15 7 at 1.5/~, resolution for a constrained model (C&S coordinates). Benzamidineinhibited bovine trypsin has also been independently refined (R = 0.229 at 1.8 ,/k) by Bode & Schwager [J. Mol. Biol. (1975), 98, 693-717] (B&S coordinate set). Comparison of these structures after suitable correction for the different inhibitors and consequent structural 0567-7408/79/081861-14501.00 differences permits an experimental determination of the differences in structure, which places an upper limit on the errors in the refined coordinate sets. The models are remarkably similar in well determined regions. The average positional difference between internal mainchain atoms is 0.146 A (0.163 A r.m.s.); however, there are some differences as large as 3-9 A associated mostly with poorly determined external side chains. The magnitude of the deviations is strongly dependent on the region of the structure compared, and is closely related to refined thermal parameters in our analysis. © 1979 International Union of Crystallography 1862 TWO INDEPENDENTLY REFINED MODELS OF BOVINE TRYPSIN Estimated errors in the C&S atom positions average 0.15-0-20 /k overall, and range from about O-10/k in well determined areas to 1 A or more in poorly determined regions. Errors in the B&S positions are expected to be somewhat larger. Comparison of structure factors computed from the B&S coordinates [F~ss )] gives a residual with the C&S observed data of R = 0.238 at 1.8/k (very close to their own residual of 0.229), indicating that the contribution from real differences between the two 'true' structures or from errors in either observed data set is small compared with errors in the refined coordinate sets. The residual between Fe~Bs ) and structure factors computed from the C&S model was 0.253 at 1.5 A, suggesting that the two models do not differ from the 'true' structure in a systematically identical fashion, and that further refinement should improve either model. The similarity of the 0.253 value to their own residual of 0.229 at 1.8 A also suggests that the differences between the C&S and B&S models are about as large as the differences between the B&S structure and the hypothetical 'true' structure. The differences between the two models thus give a reasonable estimate of the kinds of errors which can be expected in a structure with refinement statistics similar to those of Bode & Schwager.