Decoding of genome function and evolution based on the 3D structures of proteins

作为基因组序列中推断的ORF的功能预测方法，我们开发了一种将蛋白质模块的结构和功能特性转换为氨基酸序列模式的方法，称为“3D-keynote”。使用该方法，我们推断的ORF，slr 0197从蓝藻基因组的功能，这是随后通过合作实验验证。在此基础上，我们开发了一个自动生成各类函数三维特征的系统，共生成了196种三维特征，其预测精度平均达到85%左右。将这些应用于蓝藻基因组，我们获得了推断约12%的功能未知的ORF的功能的线索。3D-keynotes也被用于H-invitational（Yura & Go）中的人类基因组注释。对预测准确性的严格和宽松评估分别为25%和66%，优于其他预测方法。 ...更多信息 g方法。为了实验验证预测系统，进行了糖结合蛋白（如凝集素）的晶体结构分析。我们对6种新的congerin复合物进行了结构测定，获得了支持Shirai预测系统预测的糖结合位点存在的证据，发现疏水簇残基的链状拓扑结构是整个蛋白质分子链的简化表示。从疏水簇残基组成的假肽链的分子动力学模拟和生成结构的簇分析，我们将疏水簇残基的作用定义为它们的非特异性疏水聚集力使蛋白质链紧凑以减少构象搜索空间并加速折叠过程（Soda）。长度cDNA数据，我们开发的方法来确定选择性剪接（AS）区域，我们发现，大多数被AS改变的区域短于蛋白质结构域的常见大小，并指出转录本的AS可能导致结构不稳定和/或失去其蛋白质产物上的相互作用位点，并导致所涉及的蛋白质网络的途径改变（Go & Takahashi）。少

As a functional prediction method for ORFs inferred in genome sequences, we developed a method to convert structural and functional properties of protein modules into amino acid sequence patterns, called "3D-keynote". Using the method, we inferred a function of ORF, slr0197 from a cyanobacterium genome, which was subsequently verified by collaborative experiments. Furthermore, we made an automatic system to generate 3D-keynotes for each kind of functions and accomplished generations of 196 kinds of 3D-keynotes, whose prediction accuracies were evaluated to be about 85% on an average. Applying these to a cyanobacterium genome, we obtained clues for inferring the functions for about 12% of function-unknown ORFs. The 3D-keynotes were also used for human genome annotation in H-invitational (Yura ＆ Go).A sugar-protein interaction prediction system was developed. Strict and loose evaluations of the prediction accuracy resulted in 25 and 66%, respectively, which were better than other existin … More g methods. To experimentally verify the prediction system, crystal structure analyses of sugar-binding proteins such as lectin, were done. We determined structures of six novel complex forms of congerin and obtained evidences supporting the existence of sugar-binding sites predicted by the prediction system (Shirai).A chain topology of hydrophobic cluster residues was found to be a reduced representation of a whole chain of the protein molecule. From molecular dynamics simulations of a pseudo-peptide chain composed of hydrophobic cluster residues and cluster analyses of the generated structures, we defined roles of hydrophobic cluster residues as their non-specific hydrophobic aggregation forces made the protein chain compact to decrease the search space for conformations and accelerate the folding process (Soda).Exhaustively analyzing human full-length cDNA data by the method we developed to identify alternative splicing (AS) regions, we found that most regions altered by AS were shorter than common sizes of protein domains, and indicated possibilities that AS of transcripts may lead to structural destabilization of and/or loss of interaction sites on their protein products and result in changing pathways of the protein network involved (Go ＆ Takahashi). Less