Ogata K, Umeyama H
School of Pharmaceutical Sciences, Kitasato University, Tokyo, Japan.
Proteins. 1998 Jun 1;31(4):355-69.
We investigated the conservation of sidechain conformation for each residue within a homologous family of proteins in the Protein Data Bank (PDB) and performed sidechain modeling using this information. The information was represented by the probability of conserved sidechain torsional angles obtained from many families of proteins, and these were calculated for a pair of residues at topologically equivalent positions as a result of structural alignment. Probabilities were obtained for a pair of same amino acids and for a pair of different amino acids. The correlation between environmental residues and the fluctuation of probability was examined for the pair of same amino acid residues, and the simple probability was calculated for the pair of different amino acids. From the results on the same amino acid pairs, 17 amino acids, except for Ala, Gly, and Pro, were divided into two types: those that were influenced and those that were not influenced by the environmental residues. From results on different amino acid pairs, a replacement between large residues, such as Trp, Phe, and Tyr, was performed assuming conservation of their torsional angles within a homologous family of proteins. We performed sidechain modeling for 11 known proteins from their native and modeled backbones, respectively. With the native backbones, the percentage of the chi 1 angle correct within 30 degrees was found to be 67% and 80% for all and core residues, respectively. With the modeled backbones, the percentage of the correct chi 1 angle was found to be 60% and 72% for all and core residues, respectively. To estimate an upper limit on the accuracy for predicting sidechain conformations, we investigated the probability of conserved sidechain torsional angles for highly similar proteins having > 90% sequence identity and < 2.5-A X-ray resolution. In those proteins, 83% of the sidechain conformations were conserved for the chi 1 angle.
我们研究了蛋白质数据库（PDB）中同源蛋白家族内每个残基侧链构象的保守性，并利用这些信息进行侧链建模。该信息由从多个蛋白质家族获得的保守侧链扭转角的概率表示，这些概率是通过结构比对在拓扑等效位置的一对残基上计算得出的。计算了一对相同氨基酸和一对不同氨基酸的概率。对于一对相同氨基酸残基，研究了环境残基与概率波动之间的相关性，并计算了一对不同氨基酸的简单概率。根据相同氨基酸对的结果，除丙氨酸、甘氨酸和脯氨酸外的17种氨基酸被分为两类：受环境残基影响的和不受环境残基影响的。根据不同氨基酸对的结果，假设在同源蛋白家族内其扭转角保守，进行了色氨酸、苯丙氨酸和酪氨酸等大残基之间的替换。我们分别从11种已知蛋白质的天然骨架和建模骨架进行侧链建模。对于天然骨架，所有残基和核心残基的χ1角在30度内正确的百分比分别为67%和80%。对于建模骨架，所有残基和核心残基的正确χ1角百分比分别为60%和72%。为了估计预测侧链构象准确性的上限，我们研究了序列同一性>90%且X射线分辨率<2.5 Å的高度相似蛋白质的保守侧链扭转角的概率。在这些蛋白质中，83%的侧链构象对于χ1角是保守的。
