Division of Biostatistics, College of Public Health, The Ohio State University, United States; Department of Statistics, The Ohio State University, United States.
Center for Medical Informatics, Yale University, United States.
Mol Phylogenet Evol. 2021 Aug;161:107142. doi: 10.1016/j.ympev.2021.107142. Epub 2021 Mar 11.
Despite the recent availability of large-scale genomic data for many individuals, few methods for phylogenetic inference are both computationally efficient and highly accurate for trees with hundreds of taxa. Model-based methods such as those developed in the maximum likelihood and Bayesian frameworks are especially time-consuming, as they involve both computationally intensive calculations on fixed phylogenies and searches through the space of possible phylogenies, and they are known to scale poorly with the addition of taxa. Here, we propose a fast approximation to the maximum likelihood estimator that directly uses continuous trait data, such as allele frequency data. The approximation works by first computing the maximum likelihood estimates of some internal branch lengths, and then inferring the tree-topology using these estimates. Our approach is more computationally efficient than existing methods for such data while still achieving comparable accuracy. This method is innovative in its use of the mathematical properties of tree-topologies for inference, and thus serves as a useful addition to the collection of methods available for estimating phylogenies from continuous trait data.
尽管最近有大量的个体基因组数据可供使用，但对于具有数百个分类群的系统发育树，很少有计算效率高且高度准确的系统发育推断方法。基于模型的方法，如最大似然法和贝叶斯框架中开发的方法，特别耗时，因为它们既涉及固定系统发育树上的计算密集型计算，又涉及可能系统发育空间的搜索，并且已知随着分类群的增加而扩展不佳。在这里，我们提出了一种快速逼近最大似然估计量的方法，该方法可以直接使用连续特征数据，例如等位基因频率数据。该逼近方法的工作原理是首先计算某些内部分支长度的最大似然估计量，然后使用这些估计量推断树拓扑结构。与现有方法相比，我们的方法在处理此类数据时更具计算效率，同时仍能达到相当的准确性。该方法在利用拓扑结构的数学性质进行推断方面具有创新性，因此是从连续特征数据估计系统发育的可用方法集合中的有用补充。
