Collaborative Research: Estimation of Large Species/Population Trees Using Tree Space

基于包含在生物体DNA序列中的信息来估计生物体的进化史是进化生物学中一个非常重要的问题。基因组测序项目产生的大量DNA序列数据在估计这些系统发育关系方面带来了重要的计算挑战。这些挑战之一是基于基因组中几个不同基因的DNA序列信息来估计一组物种的进化史。本研究的重点是在考虑的物种数量非常大（即数百到数千）时，开发计算效率高的方法来估计进化史。这是通过一次考虑三个物种的集合，并使用估计的三个群体的进化历史的特性来推断整体的进化历史来完成的。该方法的性质和性能将在理论上以及模拟和经验数据集进行评估。这项工作有许多实际应用，例如研究人类种群之间的进化关系。虽然在过去10年中，用于推断系统发育物种树的基因组数据量迅速增加，但很少有方法能够有效地估计由数百或数千种物种组成的数据集的物种树。提出了一种快速逼近最大似然估计（MLE）的方法，该方法保留了理想的统计性质，如一致性和渐近效率。初步工作的结果表明，该方法将比现有的似然和贝叶斯方法快得多，同时也具有很高的准确性。该方法可应用于一系列数据类型，包括沿系统发育布朗运动模型产生的等位基因频率数据和聚结模型产生的单核苷酸多态性（SNP）数据。将开发一个软件包来实施该方法。该项目还将支持一名博士生，他将为该方法的开发和实施做出贡献。

The estimation of the evolutionary history of a collection of organisms based on the information contained in their DNA sequences is a problem of fundamental importance in evolutionary biology. The abundance of DNA sequence data arising from genome sequencing projects has led to important computational challenges in the estimation of these phylogenetic relationships. Among these challenges is the estimation of the evolutionary history for a group of species based on DNA sequence information from several distinct genes sampled throughout the genome. This research is focused on the development of computationally efficient methods for estimating the evolutionary history when the number of species under consideration is very large (i.e., hundreds to thousands). This is accomplished by considering collections of three species at a time, and using properties of the estimated evolutionary history for groups of three to infer the overall evolutionary history. Properties and performance of the method will be evaluated theoretically as well as with both simulated and empirical data sets. This work has numerous practical applications, such as the study of the evolutionary relationships among human populations.Though the amount of genomic data available for inferring phylogenetic species trees has increased rapidly within the last 10 years, few methods have been developed to efficiently estimate species trees for data sets consisting of hundreds or thousands of species. A fast approximation to the maximum likelihood estimate (MLE) that retains desirable statistical properties, such as consistency and asymptotic efficiency, is proposed. Results from preliminary work suggest that this approach will be significantly faster than existing likelihood and Bayesian approaches, while also being highly accurate. The method can be applied to a range of data types, including allele frequency data arising under a Brownian motion model along the phylogeny and single nucleotide polymorphism (SNP) data arising from the coalescent model. A software package will be developed to implement the methodology. The project will also support one PhD student, who will contribute to the development and implementation of the methodology.