Data from: Delimiting species using single-locus data and the Generalized Mixed Yule Coalescent approach: a revised method and evaluation on simulated data sets

DNA barcoding-type studies assemble single-locus data from large samples of individuals and species, and have provided new kinds of data for evolutionary surveys of diversity. An important goal of many such studies is to delimit evolutionarily significant species units, especially in biodiversity surveys from environmental DNA samples. The Generalized Mixed Yule Coalescent (GMYC) method is a likelihood method for delimiting species by fitting within- and between-species branching models to reconstructed gene trees. Although the method has been widely used, it has not previously been described in detail or evaluated fully against simulations of alternative scenarios of true patterns of population variation and divergence between species. Here, we present important reformulations to the GMYC method as originally specified, and demonstrate its robustness to a range of departures from its simplifying assumptions. The main factor affecting the accuracy of delimitation is the mean population size of species relative to divergence times between them. Other departures from the model assumptions, such as varying population sizes among species, alternative scenarios for speciation and extinction, and population growth or subdivision within species, have relatively smaller effects. Our simulations demonstrate that support measures derived from the likelihood function provide a robust indication of when the model performs well and when it leads to inaccurate delimitations. Finally, the so-called single-threshold version of the method outperforms the multiple-threshold version of the method on simulated data: we argue that this might represent a fundamental limit due to the nature of evidence used to delimit species in this approach. Together with other studies comparing its performance relative to other methods, our findings support the robustness of GMYC as a tool for delimiting species when only single-locus information is available.

DNA条形码（DNA barcoding）类研究从大量个体与物种样本中组装单基因座数据，为多样性进化调查提供了新型数据。此类研究的一项核心目标是划定进化显著性物种单元，尤其在基于环境DNA样本的生物多样性调查中。广义混合尤尔-合并模型（Generalized Mixed Yule Coalescent, GMYC）是一种似然方法，通过将种内与种间分支模型拟合至重建的基因树来实现物种划定。尽管该方法已得到广泛应用，但此前尚未针对种群变异与物种间分化的真实模式的各类模拟情景进行详尽描述与全面评估。本文对初始设定的GMYC方法进行了重要修正，并验证了其在一系列偏离简化假设场景下的稳健性。影响划定准确性的核心因素为物种平均种群大小与其间分化时间的相对比值。其他偏离模型假设的情况，如物种间种群大小存在差异、物种形成与灭绝的不同情景、种内种群增长或分化，其影响相对较小。我们的模拟实验表明，源自似然函数的支持度指标可稳健地指示模型何时表现良好、何时会导致不准确的物种划定。最后，在模拟数据上，所谓的单阈值版本方法的表现优于多阈值版本方法，我们认为这可能是由于该方法划定物种时所依据的证据本质，造成了这一根本性局限。结合其他对比该方法与其他方法性能的研究，我们的研究结果证实，当仅能获取单基因座信息时，GMYC作为物种划定工具具有稳健性。