Department of Bio and Brain Engineering, KAIST, Daejeon 305-701, Korea.
Bioinformatics. 2009 Oct 1;25(19):2506-13. doi: 10.1093/bioinformatics/btp455. Epub 2009 Jul 23.
Diverse studies have shown that correlated mutation (CM) is an important molecular evolutionary process alongside conservation. However, attempts to find the residue pairs that co-evolve under the structural and/or functional constraints are complicated by the fact that a large portion of covariance signals found in multiple sequence alignments arise from correlations due to common ancestry and stochastic noise.
Assuming that the background noise can be estimated from the coevolutionary relationships among residues, we propose a new measure for background noise called the normalized coevolutionary pattern similarity (NCPS) score. By subtracting NCPS scores from raw CM scores and combining the results with an entropy factor, we show that these new scores effectively reduce the background noise. To test the effectiveness of this method in detecting residue pairs coevolving under the structural constraints, two independent test sets were performed, showing that this new method performs better than the most accurate method currently available. In addition, we also applied our method to double mutant cycle experiments and protein-protein interactions. Although more rigorous tests are required, we obtained promising results that our method tended to explain those data better than other methods. These results suggest that the new noise-reduced CM scores developed in this study can be a valuable tool for the study of correlated mutations under the structural and/or functional constraints in proteins.
许多研究表明，相关突变（CM）是除保守性之外的另一种重要的分子进化过程。然而，尝试寻找在结构和/或功能约束下共同进化的残基对的方法受到了以下事实的影响：在多序列比对中发现的大部分协方差信号是由于共同祖先和随机噪声引起的相关性产生的。
假设可以从残基之间的共进化关系中估计背景噪声，我们提出了一种新的背景噪声度量，称为归一化共进化模式相似性（NCPS）得分。通过从原始 CM 得分中减去 NCPS 得分，并将结果与熵因子相结合，我们表明这些新得分可以有效地降低背景噪声。为了测试该方法在检测结构约束下共同进化的残基对的有效性，我们进行了两个独立的测试集，结果表明该新方法的性能优于目前最准确的方法。此外，我们还将该方法应用于双突变体循环实验和蛋白质-蛋白质相互作用。尽管需要更严格的测试，但我们获得了有希望的结果，表明该方法比其他方法更能解释这些数据。这些结果表明，该研究中开发的新的降低噪声的 CM 得分可以成为研究蛋白质中结构和/或功能约束下相关突变的有用工具。
