Efficient computational technologies to resolve the Timetree of Life: from ancient DNA to species-rich phylogenies

生物的基因组随着时间的推移而积累变化，像钟表一样记录着物种分化和灭绝的进化史。通过对基因组进行测序并整合化石记录中的信息，科学家可以重建进化树并根据地质时间进行校准。这种分子时钟测年方法通过利用测序基因组的信息产生了惊人的结果，极大地丰富了我们对物种随时间多样化的宏观进化过程以及可能的驱动因素，如主要地质事件和古气候变化等的理解。贝叶斯推理为整合分子时钟测年分析中的各种不确定性来源提供了灵活的框架，如分子进化速度随时间的不确定性，化石在物种系统发育中的年龄和位置的不确定性，以及从基因组数据估计分子分支长度的不确定性。然而，贝叶斯方法的计算量很大，因为它们需要随机的MCMC抽样来积分不确定因素，因此它们无法处理泛滥的基因组数据。这一计算限制是一个主要问题，因为需要基因组规模的数据集来实现估计进化时间尺度的高精度，这允许测试精确的进化假设，而使用小分子数据集获得的粗略时间估计永远无法解决这些假设。我们的团队在成功开发贝叶斯方法来推断进化时间表方面有着成功的记录，我们的软件包BPP和MCMCtree被学术受益者广泛用于重建这样的时间尺度。这项提议的总体目标是开发新的模型和算法，用于使用大型基因组规模的数据集有效地推断进化时间尺度，并将这些新方法整合到我们现有的软件中。我们将开发(1)解释序列错误的模型(以适应DNA降解)并整合过时的样本用于古代基因组分析，(2)交叉支撑校准以利用基因复制和水平基因转移事件中的信息来推断古代时间尺度，以及(3)用于物种丰富的系统基因组数据集分析的高效MCMC采样算法。最后，我们将使用我们新开发的技术来处理三个重要的案例研究：(1)通过将古代和现代基因组的分析整合到多物种融合模型中并引入渐渗，(2)在过去三个主要灭绝事件(哺乳动物、鸟类、鳄鱼、海龟、鳞龙和两栖动物)的背景下，推断物种丰富的四足动物生命树(3)通过整合来自古代基因复制和水平基因转移事件的信息，推断真核生物的生命之树，这将有助于减少地球生命史上一些最深刻的物种形成事件的时间不确定性。

The genomes of organisms accumulate changes with the passage of time, recording the evolutionary history of species divergences and extinctions as a timepiece does. By sequencing genomes and integrating information from fossil records, scientists can reconstruct evolutionary trees and calibrate them to geological time. This molecular-clock dating approach has led to spectacular results by leveraging information from sequenced genomes, greatly enriching our understanding of the macroevolutionary process of species diversifications through time, and the possible driving factors such as major geological events and paleoclimate changes.Bayesian inference provides a flexible framework for integrating various sources of uncertainty in molecular clock dating analyses, such as uncertainties in the rate of molecular evolution over time, in the age and placement of fossils on species phylogenies, and in estimated molecular branch lengths from genomic data. However, Bayesian methods are computationally expensive as they require stochastic MCMC sampling to integrate over the uncertainties, and thus they are unable to handle the deluge of genomic data. This computational limitation is a major problem because genome-scale datasets are needed to achieve high precision in estimated evolutionary timescales, which allows the testing of precise evolutionary hypotheses that could never be addressed with the crude time estimates obtained with small molecular datasets.Our team has a track record of successfully developing Bayesian methods to infer evolutionary timescales, with our software packages, BPP and MCMCtree, being widely used by academic beneficiaries to reconstruct such timescales. The overarching aim of this proposal is to develop new models and algorithms for efficient inference of evolutionary timescales using large genome-scale datasets, and integrate these new methods into our existing software. We will develop (1) models that account for sequence errors (to accommodate DNA degradation) and integrate dated samples for analysis of ancient genomes, (2) cross-bracing calibrations to leverage information in gene duplications and horizontal gene transfer events for inferring ancient timescales, and (3) efficient MCMC sampling algorithms for analysis of species-rich phylogenomic datasets. The improved algorithms will make it possible to analyse large datasets with thousands of genomes, from the Darwin Tree of Life and Earth BioGenome projects, as well as helping reduce the carbon footprint of computational analyses.Finally, we will use our newly developed technologies to tackle three important case studies: (1) The evolution of human populations and other hominids, by integrating analysis of ancient and modern genomes within the multi-species coalescent model with introgression, (2) inferring the species-rich Tree of Life of Tetrapods (mammals, birds, crocodiles, turtles, lepidosaurs, and amphibians) in the context of three major past extinction events (end-Permian, Triassic-Jurassic, and end-Cretaceous), and the rapid global warming event during the Eocene-Palaeocene Thermal Maximum, and (3) inferring the Tree of Life of Eukaryotes, by integrating information from ancient gene duplication and horizontal gene transfer events, which will help reduce uncertainty in the timing some of the deepest speciation events in the history of Life on Earth.