结构与序列结合的蛋白质家族共进化网络预测的理论方法研究及在酶分子改造中的应用

The site-directed mutations of amino acids in wild enzyme are a powerful technique in protein engineering for improving the bioactivity and physicochemical properties of enzymes. For better experimental results in many cases we need the combined mutations at several residue positions. However, the possibility of combined mutations is a huge number. For example, in a protein consisting of 400 amino acids, the possible mutations in three residue positions are around 100 billion. In the long evolutionary history of enzyme family the functional residues form coevolusion network determing the evolution direction of enzyme family. To predict the mutation positions and the amino acid types is the most difficult problem in protein engineering. So far the prediction theory and methods for the coevolusion network in protein engineering were built on the statistical analysis of the history data of amino acid sequence of enzyme family, no structural data are used. We belive that the most important feature of a protein family is the common fold patern (3D structure). The structural conservation is more important than the sequence conseveation. To include the structural information into the prediction method is necessary. In this research project a computational approach, namely structure and sequence position coupling method (SSPCM), is developed to predict the coevolution network. The basic hypothesis of SSPCM is that in a protein family there is a high conservative domain in both structure and in amino acid sequence, which plays the central role of biological activity. The local structural changes may contain information of biology functional evolution. A standard protein P(0) is defined in a protein family, which consists of the most-frequent amino acids and takes the average structure of the protein family. The structural positions are organized as segments that are the stable units in structural displacements of the protein family. The foundational variables of SSPCM is the pairwise coupling free energy between position sites, which are computed using the structural data and sequence data. The biological function differences of protein members are determined by the position structural displacements of individual protein P(i) to the standard protein P(0). Correlation analysis is used to analyze the communication network among segments (positions). Based on the predited coevolution network using SSPCM new experimental stratyge is designed for the multiple site-directed mutation experiments of alpha-amylases for better experiment results and to test the prediction SSPCM theory and method.

在蛋白质家族的进化史中，不同位点的多个氨基酸形成相互关联的网络，协同变异，共同影响蛋白质功能的进化。在野生酶的氨基酸人工定点变异改造中，预测酶家族的'共进化网络'，在多个位点上做联合变异，定向改进酶的生物活性和品质，是蛋白质工程领域中世界性难题。已有方法以家族氨基酸序列统计数据为基础，用信息理论等方法预测'共进化网络'。预研发现蛋白质家族的结构保守性高于序列的保守性，我们提出结构与序列相结合的预测方法。首先建立酶家族的三维结构比对和序列比对数据集，结合序列的保守性和结构的保守性信息，找出家族最保守的核心结构区，构造家族的'标准蛋白'。计算家族成员的结构位点位移矩阵和频率偏移矩阵，进而计算偶合自由能矩阵，通过氨基酸位点间的相关系数，找出相关性高的位点间的共进化网络，提出自己的预测理论和方法。把该方法用于正在进行的Alpha-淀粉酶的生物技术改造实验，验证理论，指导实验。

从20种氨基酸间的多种相互作用的物理和化学机理及蛋白质家族的生物活性和理化品质与残基序列结构的统计关系两方面进行了系统的研究。一方面用量子化学方法深入研究了氨基酸间的几种强相互作用，包括阳离子-π作用、极性氢-π作用、盐桥作用、酰胺桥作用和疏水作用的数量等级。另一方面，用氨基酸的理化性质和蛋白质家族的氨基酸序列结构为参数，建立了2种研究蛋白质家族的生物活性和理化品质与残基序列结构关系的统计数学的预测计算方法，包括氨基酸序列和理化性质的双层构效关系（2L-QSAR），及蛋白质序列位点和氨基酸理化性质的双层主成分分析法（2L-PCA）。在双层构效关系（2L-QSAR）中建立了酶分子活性与两类参数（氨基酸理化性质和蛋白质氨基酸系列结构）的嵌套式统计预测关系。在双层主成分分析法（2L-PCA）中进一步建立了氨基酸理化性质的主成分向量的正交空间和蛋白质残基的序列位点的主成分向量的正交空间.在这两个相互嵌套的正交空间中，蛋白质家族的每一个蛋白质是正交空间里的一个向量，它在两个正交空间的主成分分量上的投影为分析两类参数对酶的活性的影响提供了数量依据。提出了改进酶分子催化反应的pH值适应性的“活性氢键网络法”和改进酶分子热稳定性的“蛋白质模拟热探测法”，把以上的理论研究成果结合起来，应用于α-淀粉酶和普鲁蓝酶分子的改造，在酶分子的热稳定性和pH的耐受性的改造方面取得了较好的结果。