An Greatly Expanded CATH-Gene3D with Functional Fingerprints to Characterise Proteins

有数以百万计的蛋白质正在被测序，这些蛋白质没有已知的功能。新的Cath方法将预测它们的功能。虽然其他资源也可以预测功能，但Cath-Gene3D(以下称为Cath)提供了与功能相关的结构保守特征的独特信息。结构数据揭示了蛋白质如何执行其功能，以及如果蛋白质被突变或其他基因变异修改，为什么功能会改变。蛋白质功能信息是理解生物系统的关键，进而是药物设计、蛋白质工程和疾病的关键。CATH是世界领先的资源，将从相同祖先蛋白质进化而来的蛋白质分类为进化家族。目前，CATH将1500万个蛋白质结构域划分为2600个家族。家族数据很有价值，因为进化中的近亲(称为同源基因)往往具有相似的3D结构和执行相似的功能。因此，CATH的好处是能够推断同系物之间的性质。这一点很重要，因为目前已知的数百万蛋白质(&gt；2000万)中，只有不到5%的蛋白质具有实验确定的功能。即使在我们最感兴趣的生物体中，人类也有10%的蛋白质具有已知的功能。由于确定蛋白质的特性可能会很慢，而且成本很高，所以不可能对所有这些蛋白质进行实验研究。因此，生物学家使用CATH来根据蛋白质所属的家族来预测蛋白质的功能。另一个事实是蛋白质是由“结构域”组成的--平均每个蛋白质有两个结构域。这些是独立折叠的实体，它们共同作用，赋予整个蛋白质的功能。CATH在结构域的水平上对蛋白质进行分类，目前对自然界中发现的~70%的结构域进行了分类。结构域是蛋白质的基石--几千个结构域以不同的方式结合在一起，形成了自然界中的2000万个或更多的蛋白质。我们小组开发了预测领域函数的方法。这使得整个蛋白质的功能可以从它们的构成结构域的功能中推断出来。CATH使用结构域的三维结构信息来给出更准确的家族分类，因为结构在进化过程中比序列更高度保守。更重要的是，结构可以揭示蛋白质如何执行其功能，以及如果特定部位发生突变，蛋白质是否会失去功能。我们将把Cath扩展100%。由于人工验证非常耗时，我们将开发更好的方法来自动识别远距离同源。我们将在人工筛选之前持续发布数据(Cath-B)，以便生物学家能够更快地从这些信息中受益。我们将与其他主要结构分类SCOP合作，制定共同的分类策略，并提供关于家族的补充信息。我们将提高跨家族功能遗传的准确性。我们需要这样做，因为在一些家族中，特别是那些在自然界中出现更频繁的家族，功能可能会在某些亲属中发生变化。我们将通过表征领域中的重要位置来提高准确性，这些位置在功能相似的亲属中保守。我们可以建立这些位置的模式，以识别共享这些模式并可能具有类似功能的其他结构域。我们将使生物学家更容易使用我们的网络搜索工具来确定蛋白质是否属于这些功能家族之一。我们将把它设置在云上，这样生物学家就可以用他们使用新的测序技术获得的海量数据集快速搜索Cath。这些技术捕捉在不同条件下表达的蛋白质。我们的网页将报告它们的功能和蛋白质的变异，这可能会改变导致疾病的功能

There are millions of proteins being sequenced which have no known function. New CATH methods will predict their functions. Whilst other resources do also predict function, CATH-Gene3D (referred to below as CATH) provides unique information on structurally conserved features linked to function. Structure data reveals how proteins perform their function and why the function changes if the protein is modified by mutations or other genetic variations. Protein function information is key to understanding biological systems and by extension drug design, protein engineering and disease.CATH is a world leading resource that classifies proteins evolved from the same ancestral protein, into evolutionary families. Currently, CATH classifies 15 million protein domains into 2600 families. Family data is valuable because evolutionary relatives (called homologues) tend to have similar 3D structures and perform similar functions. Thus the benefit of CATH is the ability to infer properties between homologues.This is important because of the millions of proteins currently known (>20 million) less than 5% have experimentally determined functions. Even in the organism of greatest interest to us, human, <10% of proteins have known functions. Because it can be slow and very expensive to characterise proteins it will not be possible to experimentally study all these proteins. Therefore, biologists use CATH to predict the function of a protein based on the family to which it belongs.Another fact is that proteins are made up of 'domains' - on average two per protein. These are independently folded entities that act together to confer the function of the whole protein. CATH classifies proteins at the level of the domain and currently classifies ~70% of domains found in nature. Domains are the building blocks of proteins - a few thousand of them are combined in different ways to give the 20 million proteins, or more, in nature. Our group develops methods for predicting domain functions. This allows functions of whole proteins to be deduced from the functions of their constitutive domains. Thus functions can be suggested for proteins made from any combination of domains.CATH uses information on the 3D structure of the domain to give more accurate family classifications, as structure is more highly conserved, during evolution, than the sequence. Even more important - structure can reveal how the protein performs its function and whether the protein loses its function if a mutation occurs at a particular site.We will expand CATH by 100%. Since manual validation is very time consuming, we will develop better methods for automatically recognising distant homologues. We will continuously release data (CATH-B), prior to manual curation, so that biologists can benefit from the information much sooner.We will collaborate with the other major structure classification SCOP to develop common classification strategies and provide complementary information on families.We will improve the accuracy of functional inheritance across a family. We need to do this because in some families, especially those occurring more frequently in nature, the functions can change in some relatives.We will improve accuracy by characterising important positions in the domain, conserved across functionally similar relatives. We can build patterns of these positions to recognise other domains sharing such patterns and likely to have similar functions.We will make it easy for biologists to use our web search tool to determine if a protein belongs to one of these functional families. We will set this up on the Cloud so that biologists can quickly search CATH with the massive datasets they obtain using new sequencing technologies. These technologies capture proteins expressed under different conditions. Our web pages will report their functions and variations in the protein which could modify function causing disease