Gatchell D W, Dennis S, Vajda S
Department of Biomedical Engineering, Boston University, Boston, Massachusetts 02215, USA.
Proteins. 2000 Dec 1;41(4):518-34.
Free energy potentials, combining molecular mechanics with empirical solvation and entropic terms, are used to discriminate native and near-native protein conformations from slightly misfolded decoys. Since the functional forms of these potentials vary within the field, it is of interest to determine the contributions of individual free energy terms and their combinations to the discriminative power of the potential. This is achieved in terms of quantitative measures of discrimination that include the correlation coefficient between RMSD and free energy, and a new measure labeled the minimum discriminatory slope (MDS). In terms of these criteria, the internal energy is shown to be a good discriminator on its own, which implies that even well-constructed decoys are substantially more strained than the native protein structure. The discrimination improves if, in addition to the internal energy, the free energy expression includes the electrostatic energy, calculated by assuming non-ionized side chains, and an empirical solvation term, with the classical atomic solvation parameter model providing slightly better discrimination than a structure-based atomic contact potential. Finally, the inclusion of a term representing the side chain entropy change, and calculated by an established empirical scale, is so inaccurate that it makes the discrimination worse. It is shown that both the correlation coefficient and the MDS value (or its dimensionless form) are needed for an objective assessment of a potential, and that together they provide much more information on the origins of discrimination than simple inspection of the RMSD-free energy plots.
结合分子力学与经验溶剂化和熵项的自由能势，用于区分天然和接近天然的蛋白质构象与轻微错误折叠的诱饵构象。由于这些势的函数形式在该领域内各不相同，因此确定各个自由能项及其组合对势的区分能力的贡献是很有意义的。这是通过定量的区分度量来实现的，这些度量包括均方根偏差（RMSD）与自由能之间的相关系数，以及一种新的度量，称为最小区分斜率（MDS）。根据这些标准，内能本身就是一个很好的区分指标，这意味着即使是构建良好的诱饵构象也比天然蛋白质结构的应变大得多。如果自由能表达式除了内能之外，还包括通过假设非离子化侧链计算的静电能和一个经验溶剂化项，那么区分能力会有所提高，其中经典原子溶剂化参数模型提供的区分能力略优于基于结构的原子接触势。最后，包含一个通过既定经验尺度计算的代表侧链熵变的项是如此不准确，以至于它会使区分能力变差。结果表明，对于势的客观评估，相关系数和MDS值（或其无量纲形式）都是必需的，而且它们一起提供的关于区分起源的信息比简单检查RMSD-自由能图要多得多。
