A Bayesian statistical approach to determine whether genetic data delimits species versus populations

最近的技术进步所提供的前所未有的DNA序列数据正在改变生物学家识别物种的方式。这些数据具有很大的力量来揭示物种之间的界限。然而，随着序列数据量的增加，检测到的遗传差异不仅与物种边界有关，还包括物种内种群之间的遗传差异。因此，物种边界被错误识别，这对整个生物学产生了深远的影响，因为物种是构建生物学问题的基本参考单位。该项目数据将开发分析方法，以避免将物种的边界与其中的当地种群结构混为一谈。它将使用计算机模拟来评估影响物种检测准确性的关键数据属性，以及新方法的性能，特别是当不同过程导致新物种形成时。这些模拟将通过自然界的例子来提供信息，确保生物现实，而不仅仅是理论理想，被共同考虑。一个由具有互补技能的研究人员组成的多元化团队将推动科学目标和外联活动，范围从为公众和决策者开发教育工具到培训参与生物多样性研究的学生。目前基于遗传学的物种划界方法可能会导致生物多样性的大规模高估，将种群和物种的差异视为统计学上的等价物。拟议的研究将解决这些限制。根据新的建模方法的物种界定，第一次物种形成被建模为一个扩展的过程，而不是被视为一个瞬时事件，因此，这种方法可以用来获得有关多样化过程本身的见解。具体来说，新方法将耦合多物种的聚结与不同的多样化模型贝叶斯统计推断，以避免混淆人口和物种的独特性。该方法将在一个免费软件包DELIMIT中传播，并参考现有的经验数据，特别是澳大利亚有鳞目动物的基因组数据集进行开发。这些属中的许多属在目前公认的物种中的数千个位点上显示出非常深的物种地理结构，这表明物种多样性被大大低估，而其他属则代表了最近的适应性辐射，因此经验系统跨越了一系列的物种形成过程。这种情况下，将被用来评估如何强大的推断物种边界是不同的多样化过程，但也验证推断DELIMIT关于物种形成过程本身的测试与非遗传信息的物种形成过程的一般对应关系。

The unprecedented amount of DNA sequence data made available by recent technological advances is changing how biologists identify species. Such data have great power to reveal the boundaries separating species. However, with increased amounts of sequence data, the genetic differences that are detected are not just associated with species boundaries, but include genetic differences among populations within species. As a consequence, species boundaries are misidentified, which has profound implications across biology because species are the basic unit of reference for framing biological questions. This project data will develop analytical methods to avoid conflating the boundaries of species with the local population structure within them. It will use computer simulations to evaluate key data properties affecting the accuracy of species detection, as well as the performance of the new method, especially when different processes give rise to the formation of new species. These simulations will be informed by examples from nature, assuring that biological realities, not just theoretical ideals, are jointly considered. A diverse team of researchers with complementary skill sets will advance both the scientific goals and outreach activities, which range from developing educational instruments for the public and policy makers to the training of students involved in biodiversity research. Current genetic-based species delimitation methods can potentially lead to mass overestimates of biodiversity by treating population and species divergence as statistically equivalent. The proposed research will address these limitations. Under the new modeling approach for species delimitation, for the first time speciation is modeled as an extended process, as opposed to being treated as an instantaneous event, and as such, this approach can be used to gain insights about the diversification process itself as well. Specifically, the new approach will couple the multispecies coalescent with different diversification models for Bayesian statistical inference to avoid conflating population and species distinctiveness. The approach will be disseminated in a free software package DELIMIT and developed with reference to existing empirical data, specifically, genomic datasets of Australian squamates. Many of these genera show extraordinarily deep phylogeographic structure across thousands of loci within currently recognized species, suggesting substantial underestimation of species diversity, whereas other genera represent recent adaptive radiations, such that the empirical systems span a range of speciation processes. This context will be used to assess how robust inferred species boundaries are to different diversification processes, but also validate inferences from DELIMIT regarding the speciation process itself by testing for a general correspondence with non-genetic information about the speciation process.