Next generation comparative methods

了解进化关系以及物种的特征(如行为、基因组、形态特征和蛋白质)如何随时间进化是所有生物学家直接或间接的基本追求。研究物种特征如何随时间变化的计算工具被称为比较方法。在其他方面，比较方法被用来重建祖先的形式，计算特征随时间变化的快慢，并测试物种特征的进化是否相互关联。来自所有研究领域的生物学家每年都会在科学出版物中数以千计地使用比较方法。分子测序技术和计算机能力的最新进展已经产生了大量非常详细的物种相互关联的地图。这些图谱以树的形式表示，类似于家谱，它们被称为系统发育树或系统发育树。系统发育树结合物种特征和比较方法来帮助生物学家推断进化的历史过程。2013年发表了两个最大的系统发育图，包括近10,000个物种的近乎完整的鸟类系统发育图和8,000个物种的大型鱼类系统发育图。这些物种加入了哺乳动物系统发育图5,000种(2007年)、55,000种植物树(2009年)和两栖动物6,000种系统发育图(2012年)。相比之下，在本世纪初，100-200个物种的系统发育被认为是非常大的。虽然数据和计算能力在过去20年里取得了惊人的进步，但大多数比较方法分析中使用的基本统计数据却跟不上步伐。统计框架是在一棵30个树种的大树被认为是大的时候建立的。这意味着，绝大多数比较方法都假定进化过程在时间和树上都是恒定的和同质的。这一假设在最初提出时并不是不合理的，因为现有的系统发育由少数密切相关的分类群组成，这些分类群涵盖的时间很短。今天，可用的系统发育图的规模已经极大地增长，它们现在涵盖了更多不同的群体和更大的时间框架，并包括了一个全面的物种样本。使用这些树，我们现在可以看到，同质假设已经被证明产生了不正确的结果，并隐藏了重要的进化信息。考虑到哺乳动物身体大小的进化，传统的比较方法假设了一个经过数亿年的同质进化过程，在所有时间段都相同地影响所有物种。但影响某些群体的进化过程已经被证明是截然不同的，例如，蝙蝠的飞行限制了它们的体型，而水生动物则允许体型增加。同质过程的假设产生了平均效应，无法检测到进化过程中的重要变化，并产生已知错误的结果。这个项目将开发新的统计方法，消除跨越系统发育树和时间的同质进化过程的假设。这些方法不仅将更准确地模拟异质进化过程，而且同样重要的是，它们能够在不事先知道的情况下自动检测这些转变的数量和位置。自动检测进化过程中的变化的能力提供了宝贵的生物学见解，使研究人员能够比以前更精细地了解进化过程。这些方法将使使用比较方法的数千名研究人员直接受益，并弥合数据进步与分析方法之间的差距。

Understanding evolutionary relationships and how characteristics of species (e.g. behaviours, genomes, morphological characteristics and proteins) evolve over time is a fundamental pursuit, either directly or indirectly, for all biologists. Computational tools to study how species characteristics change over time are called comparative methods. Among other things comparative methods are used to reconstruct ancestral forms, calculate how fast (or slow) characteristics change through time and to test if the evolution of species characteristics are correlated. Comparative methods are used thousands of time each year in scientific publications by biologists from all research areas. Recent advances in molecular sequencing technology and computer power have produced large and highly detailed maps of how species are related to each other. These maps are represented in a tree like form analogous to a family tree, they are known as phylogenies or phylogenetic trees. Phylogenetic trees are used in conjunction with species characteristics and comparative methods to help biologists infer historical processes of evolution.In 2013 two of the largest phylogenies were published, a near complete phylogeny of birds, comprising of almost 10,000 species and a large fish phylogeny of 8,000 species. These join a mammal phylogeny 5,000 species (2007), a 55,000 species tree of plants (2009) and a 6,000 species phylogeny of amphibians (2012). In contrast in the early 2000s a phylogeny of 100-200 species was considered very large. While the data and computing power have advanced inordinately over the last 20 years, the underlying statistics used in most comparative methods analysis has failed to keep pace. The statistical framework was laid down when a 30 species tree were considered large. This means that the vast majority of comparative methods assume that evolutionary processes are constant and homogeneous through time and through the tree. This assumption was not unreasonable when first introduced, as the available phylogenies consisted of a small number of closely related taxa which covered a narrow time period. Today the size of available phylogenies have grown enormously, they now cover more divergent groups and larger time frames and include a comprehensive sample of species. Using these trees we can now see that the homogenous assumption has been shown to produce incorrect results and hides important evolutionary information. Consider the evolution of body size in mammals, traditional comparative methods assume a homogeneous evolutionary process over hundreds of millions of years, affecting all species, at all time periods the same. But the evolutionary processes affecting some groups have been shown to be radically different, for example, flight in bats limit their body size, while being aquatic allows body size to increase. The assumption of a homogeneous process creates an averaging effect which is unable to detect important changes in evolutionary processes and produces results which are known to be wrong. This project will develop novel statistical methods which remove the assumption of a homogenous evolutionary process across the phylogenetic tree and through time. The methods will not only more accurately model heterogeneous evolutionary processes but of equal importance is their ability to automatically detect, without prior knowledge, the number and location of these shifts. The ability to automatically detect changes in evolutionary processes provides valuable biological insights allowing researchers to understand evolutionary processes on a finer scale than previously possible. These methods will directly benefit the thousands of researchers using comparative methods and bridge the gap between advances in data and the methods used to analyse them.