Simmons Mark P, Randle Christopher P, Freudenstein John V, Wenzel John W
Department of Evolution, Ecology, and Organismal Biology, Ohio State University, Columbus, OH, USA.
Mol Biol Evol. 2002 Jan;19(1):14-23. doi: 10.1093/oxfordjournals.molbev.a003978.
In this paper we use hypothetical and empirical data matrices to evaluate the ability of relative apparent synapomorphy analysis (RASA) to measure phylogenetic signal, select outgroups, and identify terminals subject to long-branch attraction. In all cases, except for equal character-state frequencies, RASA indicated extraordinarily high levels of phylogenetic information for hypothetical data matrices that are uninformative regarding relationships among the terminals. Yet, regardless of the number of characters or character-state frequencies, RASA failed to detect phylogenetic signal for hypothetical matrices with strong phylogenetic signal. In our empirical example, RASA indicated increasing phylogenetic signal for matrices for which the strict consensus of the most parsimonious trees is increasingly poorly resolved, clades are increasingly poorly supported, and for which many relationships are in conflict with more widely sampled analyses. RASA is an ineffective approach to identify outgroup terminal(s) with the most plesiomorphic character states for the ingroup. Our hypothetical example demonstrated that RASA preferred outgroup terminals with increasing numbers of convergent character states with ingroup terminals, and rejected the outgroup terminal with all plesiomorphic character states. Our empirical example demonstrated that RASA, in all three cases examined, selected an ingroup terminal, rather than an outgroup terminal, as the best outgroup. In no case was one of the two outgroup terminals even close to being considered the optimal outgroup by RASA. RASA is an ineffective means of identifying problematic long-branch terminals. In our hypothetical example, RASA indicated a terminal as being a problematic long-branch terminal in spite of the terminal being on a zero-length branch and having no possibility of undergoing long-branch attraction with another terminal. RASA also failed to identify actual problematic long-branch terminals that did undergo long-branch attraction, but only after following Lyons-Weiler and Hoelzer's (1997) three-step process to identify and remove terminals subject to long-branch attraction. We conclude that RASA should not be used for any of these purposes.
在本文中，我们使用假设数据矩阵和实证数据矩阵来评估相对表观共衍征分析（RASA）在测量系统发育信号、选择外类群以及识别易受长枝吸引影响的终端方面的能力。在所有情况下，除了字符状态频率相等的情况外，对于那些在终端之间关系上无信息的假设数据矩阵，RASA显示出极高水平的系统发育信息。然而，无论字符数量或字符状态频率如何，对于具有强系统发育信号的假设矩阵，RASA都未能检测到系统发育信号。在我们的实证示例中，对于最简约树的严格合意越来越难以解析、分支支持越来越差且许多关系与更广泛采样分析相冲突的矩阵，RASA显示出系统发育信号增加。RASA是一种无效的方法，无法识别具有对内类群而言最原始字符状态的外类群终端。我们的假设示例表明，RASA更倾向于选择与内类群终端具有越来越多趋同字符状态的外类群终端，而拒绝具有所有原始字符状态的外类群终端。我们的实证示例表明，在所研究的所有三种情况下，RASA选择的最佳外类群是一个内类群终端，而不是外类群终端。在任何情况下，两个外类群终端中都没有一个被RASA认为接近最佳外类群。RASA是识别有问题的长枝终端的无效手段。在我们的假设示例中，RASA将一个终端指示为有问题的长枝终端，尽管该终端位于零长度分支上，并且不可能与另一个终端发生长枝吸引。RASA也未能识别出实际发生长枝吸引的有问题的长枝终端，只是在遵循莱昂斯 - 韦勒和霍尔泽（1997）的三步过程来识别和去除易受长枝吸引影响的终端之后才识别出来。我们得出结论，RASA不应用于任何这些目的。
