Approaches for Assessing Phylogenetic Accuracy
Approaches for Assessing Phylogenetic Accuracy
复制标题
DOI：
10.1093/sysbio/44.1.3
复制
发表时间：
1995-03
影响因子：
6.5
通讯作者：
D. Hillis
中科院分区：
文献类型：
--
作者：
D. Hillis
系统发育方法的准确性可以从一致性、效率和稳健性方面进行评估。评估系统发育准确性的主要方法有四种：模拟、已知系统发育、统计分析和一致性研究。模拟研究有助于研究理想条件下方法的准确性，并可用于做出一般预测。如果考虑到模型的局限性，则可以了解方法的行为。对已知系统发育的研究可用于测试模拟研究的预测，从而检查模型的稳健性（并可能为未来的模拟提出改进建议）。统计分析允许将一般预测应用于特定结果，有助于评估是否收集了足够的数据来得出可靠的结论，并表明给定的数据集是否比随机噪声更具结构化性。最后，多个数据集的一致性研究可用于评估独立结果一致的程度，从而评估可归因于潜在系统发育的研究结果的最小比例。这些有区别吗？评估系统发育准确性的方法在很大程度上是互补的，并且结果是一致的。有助于识别一大类适合系统发育重建的问题。 [物理？洛根尼；准确性;模拟；实验进化；统计数据;一致性；一致性;效率; ] 系统发育分析已经变得com？整个生物光盘中普遍存在？过去几十年的线路。这个在?对进化史的日益重视是这一目标实现的直接结果。理解系统发育背景作为解释几乎所有生物系统的先决条件的重要性？主动语境。然而，增加尿路感染？系统发育方法的发展至少同样受到技术的推动？卡尔和方法论的进步如何？观念上的进步。特别是算法开发、计算机技术的进步？ ogy和分子生物学创造了phy？遗传分析是否适用于几乎所有涉及生物谱系的问题，例如现存人类群体中的病毒流行？ （例如，Ou 等，1992）到最早生命谱系的起源（例如，Olsen，1987）。系统发育应用依赖于系统发育树的准确重建；因此，系统学家很自然地想知道他们重建的树的准确性。本期Sys?主题生物学包含对系统学家探索的四种方法的评论，以检查系统发育的准确性和as1电子邮件：hillis@bull.zo.utexas.edu。对他们的结果充满信心：进化？元模拟（Huelsenbeck，1995），爆炸？已知（观察到的）系统发育的比率（此处简要回顾），统计评估？系统蒸发散（Li 和 Zharkikh，1995），并且 congru？科学研究（Miyamoto 和 Fitch，1995）。我评估系统发育的准确性，有两个共同的目标：人们可能会询问系统发育方法的一般属性或特定的系统发育es？亲密的。系统学家可能会解决给定方法在不同环境下的工作效果如何？环境（例如，不同的进化条件、不同的信息量、不同类型的树木等）。此类研究可能会询问系统发育表现？单个方法的效果或可以比较几种方法。这种方法主要涉及数值模拟和研究。已知系统发育的门控。另一方面，统计和一致性研究倾向于解决系统发育的具体问题？遗传准确性，即对特定系统发育研究的置信度有多大？苏丹？当然，区别并不总是很清楚，因为模拟可以用来广告？对特定经验结果的信心（例如，Hillis 等人，1994a）以及关于系统发育方法相对准确性的一般结论可以来自统计吗？逻辑或一致性研究（例如，Penny 等
?Accuracy of phylogenetic methods may be assessed in terms of consistency, efficiency, and robustness. Four principal methods have been used for assessing phylogenetic accuracy: simulation, known phylogenies, statistical analyses, and congruence studies. Simulation studies are useful for studying accuracy of methods under idealized conditions and can be used to make general predic? tions about the behavior of methods if the limitations of the models are taken into account. Studies of known phylogenies can be used to test predictions from simulation studies, thus providing a check on the robustness of the models (and possibly suggesting refinements for future simulations). Statistical analyses allow general predictions to be applied to specific results, facilitate assessments as to whether or not sufficient data have been collected to formulate a robust conclusion, and indicate whether a given data set is any more structured than random noise. Finally, congruence studies of multiple data sets can be used to assess the degree to which independent results agree and thus the minimum proportion of the findings that can be attributed to an underlying phylogeny. These differ? ent methods of assessing phylogenetic accuracy are largely complementary, and the results are consis? tent in identifying a large class of problems that are amenable to phylogenetic reconstruction. [Phy? logeny; accuracy; simulations; experimental evolution; statistics; congruence; consistency; efficiency; robustness.] Phylogenetic analyses have become com? monplace throughout the biological disci? plines during the past few decades. This in? creased emphasis on evolutionary history is a direct result of the realization of the im? portance of understanding phylogenetic background as a prerequisite to interpreting virtually any biological system in a compar? ative context. However, the increased uti? lization of phylogenetic approaches has been driven at least as much by technologi? cal and methodological advances as by con? ceptual advances. In particular, advances in algorithm development, computer technol? ogy, and molecular biology have made phy? logenetic analyses feasible for almost any problem involving biological lineages, from viral epidemics in extant human popula? tions (e.g., Ou et al., 1992) to the origins of the earliest lineages of life (e.g., Olsen, 1987). Phylogenetic applications depend on accurate reconstructions of phylogenetic trees; therefore, it is natural that systematists should wonder about the accuracy of their reconstructed trees. This issue of Sys? tematic Biology contains reviews of the four approaches that systematists have explored to examine phylogenetic accuracy and as1 E-mail: hillis@bull.zo.utexas.edu. sess confidence in their results: evolution? ary simulations (Huelsenbeck, 1995), explo? ration of known (observed) phylogenies (briefly reviewed here), statistical evalua? tions (Li and Zharkikh, 1995), and congru? ence studies (Miyamoto and Fitch, 1995). I evaluating phylogenetic accuracy, there are two common goals: one may ask about general properties of phylogenetic methods or about a specific phylogenetic es? timate. A systematist may address how well a given method works under different cir? cumstances (e.g., different evolutionary conditions, different amounts of informa? tion, different types of trees, etc.). Such studies may ask about phylogenetic perfor? mance of a single method or may compare several methods. This approach primarily involves numerical simulations and investi? gation of known phylogenies. On the other hand, statistical and congruence studies tend to address specific questions of phylo? genetic accuracy, i.e., how much confidence can be placed in a specific phylogenetic re? sult? Of course, the distinction is not always clear because simulations can be used to ad? dress confidence in a particular empirical result (e.g., Hillis et al., 1994a) and general conclusions about the relative accuracy of phylogenetic methods can come from statis? tical or congruence studies (e.g., Penny et