Robust and accurate prediction of residue-residue interactions across protein interfaces using evolutionary information.
Robust and accurate prediction of residue-residue interactions across protein interfaces using evolutionary information.
复制标题
DOI：
10.7554/elife.02030
复制
发表时间：
2014-05-01
期刊：
影响因子：
7.7
通讯作者：
Baker D
中科院分区：
文献类型：
--
作者：
Ovchinnikov S;Kamisetty H;Baker D
在进化过程中，在蛋白质界面之间进行接触的残基的氨基酸序列同一性是否是共变的？如果是这样的话，这种协方差可以用来预测跨界面的接触，并组装生物复合体的模型。我们发现，残基对确定使用伪似然为基础的方法，以协变跨蛋白质-蛋白质界面的50 S核糖体单位和28个额外的细菌蛋白质复合物与已知的结构几乎总是在复杂的接触，只要对齐序列的数量大于两种蛋白质的平均长度。我们使用这种方法，使亚基接触预测的额外的36个蛋白质复合物与未知结构，目前的模型的基础上，这些预测的三方ATP-独立的周质（TRAP）转运，三方外排系统，丙酮酸甲酸裂解酶激活酶复合物，蛋氨酸ABC转运。DOI：http://dx.doi.org/10.7554/eLife.02030.001蛋白质被认为是生命的“主力分子”，它们参与了细胞所做的几乎所有事情。蛋白质是折叠成特定三维形状的氨基酸串。蛋白质必须有正确的形状才能正常工作，因为它们通常通过与其他蛋白质或分子结合来工作，就像一把钥匙插入一把锁。因此，研究蛋白质的结构可以为蛋白质如何发挥作用提供重要的见解。两个或多个蛋白质可以结合在一起，形成复合物来执行各种任务;即使蛋白质亚基的结构已知，解决这些复合物的结构也是具有挑战性的。现在，Ovchinnikov，Kamisetty和Baker已经开发出一种方法来预测蛋白质的哪些部分在两蛋白质复合物中相互接触。不同的物种可以有相同蛋白质的拷贝;但是来自一个物种的拷贝与来自另一个物种的相关拷贝相比，在某些位置可能具有不同的氨基酸。因此，当比较来自不同物种的相互作用蛋白质对时，两种蛋白质中将存在许多不同的位置。然而，如果一个蛋白质中某个位置的氨基酸（我们称之为“X”）发生变化，另一个蛋白质中的氨基酸（比如位置“Y”）也发生变化，因此对于位置Y上的任何给定氨基酸，通常在位置X上有一个特定的氨基酸;位置X和Y被称为“共变”。Ovchinnikov等人注意到，当一对氨基酸（两种蛋白质复合物中的每种蛋白质中的一种）共同变化时，这两种氨基酸倾向于在蛋白质-蛋白质界面处相互接触。Ovchinnikov等人使用新方法预测了细菌中发现的28种蛋白质复合物中的蛋白质-蛋白质界面，并预测了细菌核糖体中蛋白质亚基之间的界面。当将这些预测与事先已知的实际结构进行比较时，发现如果所比较的每种蛋白质的拷贝数大于两种蛋白质的平均长度，则它们是准确的。Ovchinnikov等人继续预测了另外36种结构未知的细菌蛋白质复合物的蛋白质-蛋白质界面上的氨基酸，并为4种更大的复合物提供了模型。下一个挑战是将该方法扩展到仅在真核生物中发现的蛋白质复合物（即，不是细菌）。由于真核生物蛋白质的相关拷贝数往往较小，因此可比较的蛋白质较少，因此在发生“共变”时更难检测到。DOI：http://dx.doi.org/10.7554/eLife.02030.002网站
Do the amino acid sequence identities of residues that make contact across protein interfaces covary during evolution? If so, such covariance could be used to predict contacts across interfaces and assemble models of biological complexes. We find that residue pairs identified using a pseudo-likelihood-based method to covary across protein–protein interfaces in the 50S ribosomal unit and 28 additional bacterial protein complexes with known structure are almost always in contact in the complex, provided that the number of aligned sequences is greater than the average length of the two proteins. We use this method to make subunit contact predictions for an additional 36 protein complexes with unknown structures, and present models based on these predictions for the tripartite ATP-independent periplasmic (TRAP) transporter, the tripartite efflux system, the pyruvate formate lyase-activating enzyme complex, and the methionine ABC transporter. DOI: http://dx.doi.org/10.7554/eLife.02030.001 Proteins are considered the ‘workhorse molecules’ of life and they are involved in virtually everything that cells do. Proteins are strings of amino acids that have folded into a specific three-dimensional shape. Proteins must have the correct shape to function properly, as they often work by binding to other proteins or molecules—much like a key fitting into a lock. Working out the structure of a protein can, therefore, provide major insights into how the protein does its job. Two or more proteins can bind together and form a complex to perform various tasks; and solving the structures of these complexes can be challenging, even if the structures of the protein subunits are known. Now, Ovchinnikov, Kamisetty, and Baker have developed a method for predicting which parts of the proteins make contact with each other in a two-protein complex. Different species can have copies of the same proteins; but a copy from one species might have different amino acids at certain positions when compared to a related copy from another species. As such, when pairs of interacting proteins from different species are compared, there will be many positions in the two proteins that vary. However, if the amino acid at a position in one protein (let's call it ‘X’) varies, and the amino acid at, say, position ‘Y’ in the other protein also varies such that for any given amino acid at position Y there is often a specific amino acid at position X; positions X and Y are said to ‘co-vary’. Ovchinnikov et al. noticed that when a pair of amino acids (one from each protein in a two-protein complex) co-varied, these two amino acids tended to make contact with each other at the protein–protein interface. Ovchinnikov et al. used the new method to make predictions about the protein–protein interfaces in 28 protein complexes found in bacteria, and also to make a prediction about the interface between protein subunits in the bacterial ribosome. When these predictions were checked against the actual structures, which were all known beforehand, they were found to be accurate if the number of copies of each protein being compared is greater than the average length of the two proteins. Ovchinnikov et al. went on to predict the amino acids on the protein–protein interfaces for another 36 bacterial protein complexes with unknown structures, and to present models for four larger complexes. The next challenge is to extend the method to protein complexes that are found only in eukaryotes (i.e., not in bacteria). Since the number of related copies for eukaryotic proteins tends to be smaller, there are fewer proteins to compare and it is therefore harder to detect ‘covariation’ when it occurs. DOI: http://dx.doi.org/10.7554/eLife.02030.002