Brenner S E, Berry A
Cambridge Centre for Molecular Recognition, Department of Biochemistry, University of Cambridge, United Kingdom.
Protein Sci. 1994 Oct;3(10):1871-82. doi: 10.1002/pro.5560031026.
We have developed a general quantitative methodology for designing proteins de novo, which automatically produces sequences for any given plausible protein structure. The method incorporates statistical information, a theoretical description of protein structure, and motifs described in the literature. A model system embodying a portion of the quantitative methodology has been used to design many protein sequences for the phage 434 Cro and fibronectin type III domain folds, as well as several other structures. Residue sequences selected by this prototype share no significant identity with any natural protein. Nonetheless, 3-dimensional models of the designed sequences appear generally plausible. When examined using secondary structure prediction methods and profile analysis, the designed sequences generally score considerably better than the natural ones. The designed sequences are also in reasonable agreement with a sequence template. This quantitative methodology is likely to be capable of successfully designing new proteins and yielding fundamental insights about the determinants of protein structure.
我们已经开发出一种用于从头设计蛋白质的通用定量方法，该方法能自动为任何给定的合理蛋白质结构生成序列。该方法整合了统计信息、蛋白质结构的理论描述以及文献中描述的基序。一个体现部分定量方法的模型系统已被用于为噬菌体434 Cro和纤连蛋白III型结构域折叠以及其他几种结构设计许多蛋白质序列。通过这个原型选择的残基序列与任何天然蛋白质都没有显著的同源性。尽管如此，设计序列的三维模型总体上看起来是合理的。当使用二级结构预测方法和轮廓分析进行检查时，设计序列的得分通常比天然序列高得多。设计序列也与序列模板合理匹配。这种定量方法很可能能够成功设计新的蛋白质，并对蛋白质结构的决定因素产生基本的见解。
