The Arthropod Supertree of Life: An Online Interactive Resource for Testing Patterns in Arthropod Evolution and Biodiversity

大约80%的动物物种是节肢动物：包括昆虫，螃蟹和蜘蛛。从5.5亿年前的快速辐射中，它们进化到几乎占据了每一个栖息地，并利用了大多数可以想象的生活方式。今天，节肢动物支撑着几乎所有的生态群落和食物网。它们对人类具有巨大的经济和医学重要性：作为食物来源，作物害虫和疾病载体。为了了解节肢动物的生物多样性，研究它们进化的机制，并计划它们的保护，我们必须清楚地了解它们的进化关系。有成千上万的已发表的进化树，在一个较浅的水平上为特定的节肢动物群体（例如，科内的物种）以及许多试图解决更古老的分支事件。这些已发表的树木代表了极其丰富的资源，但其中大部分仍被锁定在期刊的页面中。该项目将对5,000棵或更多的节肢动物树木进行数字化，并以电子方式在线提供给所有研究人员。不幸的是，当研究人员试图比较已发表的树木时，存在严重的困难：部分原因是它们来自许多不同类型的数据（解剖学，分子，基因组和化石），部分原因是它们以更多样化的方式进行分析。更有问题的是，它们往往意味着相互矛盾的进化模式。那么，我们如何才能将所有这些信息结合在一起，产生进化生物学家和自然资源保护主义者所需要的巨大的、包罗万象的树，而不需要对数据进行挑选呢？超树方法是目前最易处理的方法，它使用客观和可重复的规则来解决冲突并找到源树之间的重叠。这种方法已经产生了有史以来最大的树木。不幸的是，超级树的构建目前仍然非常耗时和劳动密集型。此外，一旦构建，添加新树、对数据进行子采样（例如，分子或形态学），或者使用不同的方法生成超树。因此，该项目的另一个核心目标是开发一套软件工具，该软件工具将在很大程度上自动化该过程，为没有经验的用户提供在任何分类水平上为任何节肢动物组构建超树的能力（例如，种、属、科等），并且使用多个过滤标准（例如，只有最健壮的或最近的源树）。然后，我们将这些工具嵌入到包含我们数据的网站中。现有的快速超树方法并非没有问题，因此该项目的另一个关键目标将是实现和编程新的方法（新的四方连接，最大似然，保守和贝叶斯方法都在由团队成员和我们的合作者开发中）。这些新方法的特性需要表征，我们的节肢动物数据集将提供完美的测试案例，以衡量其性能。然后，我们将使用我们的超级树来询问节肢动物生物多样性研究中的一系列重要问题。哪些进化关系是很好理解的，哪些是最不确定的，需要进一步研究？哪些节肢动物群体的进化分支序列与它们作为化石出现的顺序相匹配（这些群体对校准“分子钟”很有用）？节肢动物类群的年龄与其现今的多样性之间是否存在关系？我们还将探索超级树的效用，以解决保护优先事项。在超级树的孤立分支上单独存在的物种具有高于平均水平的“进化独特性”。在这些也濒临灭绝的地方，可以提出一个强有力的理由来优先保护它们。

Around 80% of all animal species are arthropods: the group that includes insects, crabs and spiders. From their rapid radiation over 550 million years ago, they evolved to fill almost every habitat and exploit most imaginable lifestyles. Today, arthropods underpin virtually all ecological communities and food webs. They are of immense economic and medical importance to humans: as sources of food, crop pests and vectors of disease. In order to understand the biodiversity of arthropods, to investigate the mechanisms by which they evolved, and to plan for their conservation, it is vitally important that we have a clear picture of their evolutionary relationships. There are many thousands of published evolutionary trees for particular arthropod groups at a shallow level (e.g., species within families) as well as many that attempt to resolve the more ancient branching events. These published trees represent an enormously rich resource, but one that largely remains locked within the pages of journals. This project will digitise 5,000 or more trees from across the arthropods, and make them available to all researchers electronically online. Unfortunately, there are serious difficulties when researchers try to compare published trees: partly because they are derived from many different types of data (anatomy, molecules, genomes and fossils) and partly because they are analysed in an even greater variety of ways. More problematically, they often imply contradictory patterns of evolution. How, then, can we bring all of this information together to yield the giant, all-inclusive trees that evolutionary biologists and conservationists need, and do so without cherry-picking the data? Supertree methods are presently the most tractable approach, resolving conflict and finding overlap between the source trees using objective and repeatable rules. Such approaches have yielded the largest trees ever published. Unfortunately, again, the construction of supertrees is presently very time-consuming and labour-intensive. Moreover, once constructed, it is extremely difficult or impossible to add new trees, to sub-sample the data (e.g., molecules or morphology), or to generate supertrees using different methods. Another core objective of this project is therefore to develop a set of software tools that will largely automate the process, providing inexperienced users with the ability to construct a supertree for any arthropod group at any taxonomic level (e.g., species, genera, families, etc.), and using multiple filtering criteria (e.g., only the most robust or recent source trees). We will then embed these tools in the website containing our data. Existing, fast supertree methods are not without their problems, and another key objective of the project will therefore be to realise and program novel approaches (new Quartet Joining, Maximum Likelihood, Conservative and Bayesian methods are all under development by members of the team and our collaborators). The properties of these new methods need characterisation, and our arthropod dataset will offer the perfect test case against which to benchmark their performance. We will then use our supertrees to ask a range of important questions in the study of arthropod biodiversity. Which evolutionary relationships are well-understood, and which are most uncertain and in need of further research? Which arthropod groups have an evolutionary branching sequence that matches the order in which they appear as fossils (such groups are useful for calibrating 'molecular clocks')? Is there a relationship between the age of arthropod groups and their present day diversity? We will also explore the utility of supertrees for addressing conservation priorities. Species that are alone on isolated branches of the supertree have greater than average 'evolutionary distinctiveness'. Where these are also imminently endangered, a powerful case can be mounted to prioritise their preservation.